Towards a Grammar of Gumer - Phonology and Morphology of a Western Gurage Variety

Home , Ezha, Gumer, Inor language, Soddo (woreda), Tigre language, Zay language

Zurich Open Repository and Archive University of Zurich Main Library Strickhofstrasse 39 CH-8057 Zurich www.zora.uzh.ch

Year: 2017

Völlmin, Sascha

Posted at the Zurich Open Repository and Archive, University of Zurich ZORA URL: https://doi.org/10.5167/uzh-149973 Dissertation Published Version

Originally published at: Völlmin, Sascha. Towards a Grammar of Gumer - Phonology and Morphology of a Western Gurage Variety. 2017, University of Zurich, Faculty of Arts. Towards a Grammar of Gumer Phonology and Morphology of a Western Gurage Variety

Thesis presented to the Faculty of Arts and Social Sciences of the University of Zurich for the degree of Doctor of Philosophy

by Sascha Völlmin

Accepted in the fall semester 2012 on the recommendation of the doctoral committee: Prof. Dr. Balthasar Bickel (main supervisor) Prof. Dr. Walter Bisang

Zürich 2017

Contents

Acknowledgements vii

Abbreviations viii

1 Introduction 1 1.1 Scope of the thesis ...... 1 1.2 Gumer people and language ...... 1 1.3 Classification of the language ...... 3 1.3.1 Gumer vs. Chaha ...... 4 1.4 Previous studies ...... 6 1.5 Data and method ...... 6

2 Phonology 8 2.1 Consonants ...... 8 2.1.1 Distribution of β, b and p ...... 13 2.1.2 Distribution of x and k ...... 15 2.1.3 Distribution of r and n (and l) ...... 17 2.1.3.1 Initial nasalization ...... 17 2.1.3.2 ‘Geminate’ nasalization ...... 18 2.1.3.3 Nasalization in penultimate coda position . . . 18 2.1.4 Assimilation ...... 20 2.1.4.1 Assimilation of n ...... 20 2.1.4.2 Assimilation of r to n and l ...... 21 2.1.4.3 Other cases of assimilation ...... 23 2.2 Vowels ...... 23 2.2.1 The open-mid vowels ɛ and ɔ ...... 26 2.2.2 The central vowels ɨ and ə ...... 27 2.2.2.1 Realizations of ɨ /ə in combination with y /w . . 27 2.3 Syllable structure and epenthesis ...... 28 2.3.1 Syllable structure ...... 28 2.3.2 Epenthesis ...... 32 2.3.2.1 The epenthetic vowel ɨ ...... 32 2.3.2.2 Epenthesis rules ...... 33 2.3.2.3 Epenthetic vowel ɨ between words ...... 36 2.3.3 Sequences of two vowels across morpheme boundaries . . 37 2.3.4 Short note on stress ...... 40 3 Verb morphology 41 3.1 Semitic root-and-pattern morphology ...... 41 3.2 Ethiosemitic verb types ...... 42 3.3 Gumer verb types ...... 43 3.3.1 Mutations ...... 43 3.3.2 Problem of (de-)gemination in Imperfective and Jussive . 45 3.4 Main verb types ...... 49 3.4.1 Sound verbs ...... 49 3.4.1.1 Triradicals ...... 49 3.4.1.2 Quadriradicals ...... 54 3.4.2 Weak verbs ...... 55 3.4.2.1 Weak triradicals ...... 56 3.4.2.1.1 Triradicals with weak first radical . . 56 3.4.2.1.2 Triradicals with weak second radical . 58 3.4.2.1.3 Triradicals with weak third radical . . 61 3.4.2.2 Weak quadriradicals ...... 64 3.5 Special verb classes, subtypes and exceptions ...... 67 3.5.1 Monoradicals ...... 67 3.5.2 Reduplicated verbs ...... 68 3.5.2.1 Final reduplication (122 / 1233) ...... 68 3.5.2.2 Quadriradicals with total reduplication (1212) . 69 3.5.2.3 Verbs with identical first and second radical (113) 71 3.5.2.4 Frequentative (medial reduplication 1223) . . . 72 3.5.3 Special verb types and irregular verbs ...... 72 3.5.3.1 Type 1rD ...... 72 3.5.3.2 Type 1rAI ...... 73 3.5.3.3 Verbs with penultimate r and final I / U (1rI / 1rU) 74 3.5.3.4 Verbs with prefixed n- ...... 75 3.5.3.5 The verb tʼəppʷə ‘suck’ ...... 75 3.5.3.6 The verb bar ‘say’ ...... 76 3.5.3.7 The verbs aβ ‘give’ and od ‘tell’ ...... 76 3.5.3.8 The verb wər ‘go’ ...... 77 3.5.3.9 Exceptionally geminating verbs ...... 78 3.5.3.10 Verbs with initial e ...... 79 3.5.3.11 Verbs with additional initial ə ...... 79 3.5.3.12 Amharic loan verbs ...... 80 3.5.3.13 Suppletive Imperatives ...... 81 3.6 Derived stems ...... 82 3.6.1 Derived stems with t(ə)- (detransitivizer) ...... 83 3.6.1.1 Allomorphs tə- vs. t- ...... 85 3.6.1.2 Assimilation of t- ...... 87 3.6.2 Derived stems with a- (direct causativizer) ...... 88 3.6.3 Derived stems with at- (indirect causativizer) ...... 91 3.6.3.1 Assimilation of at- ...... 94

ii 3.7 Uses of t(ə)-, a- and at- ...... 94 3.7.1 Detransitivizer t(ə)-: anticausative(-passive-reflexive) . . . 95 3.7.1.1 Reciprocals ...... 97 3.7.2 Causativizer a- (direct causativization) ...... 99 3.7.3 Causativizer at- (indirect causativization) ...... 103 3.7.3.1 Causation of reciprocity ...... 108 3.8 Verbs without basic stem (prefix-necessitating or ‘bound’ stems) . 109 3.9 Note on unpredictable and specialized meanings ...... 111 3.10 Frequentatives ...... 112 3.10.1 Frequentatives of basic stems ...... 113 3.10.2 Frequentatives of derived stems ...... 116 3.11 Subject marking ...... 120 3.11.1 Perfective conjugation ...... 121 3.11.2 Imperfective conjugation ...... 122 3.11.3 Jussive conjugation ...... 122 3.11.4 Formation of feminine singular ...... 123 3.11.5 Formation of Impersonal ...... 128 3.11.6 Allomorphs of subject markers ...... 133 3.11.6.1 Verbs with weak final radical ...... 133 3.11.6.2 Verb bases with initial vowel ...... 138 3.11.6.3 1s Imperfective with prefixes ...... 139 3.11.6.4 Subject markers followed by object markers . . 140 3.12 Object marking ...... 141 3.12.1 Forms ...... 141 3.12.1.1 Fused subject-object markers ...... 143 3.12.2 Use of the object markers ...... 147 3.12.2.1 Primary object ...... 148 3.12.2.2 Benefactive and malefactive ...... 149 3.13 Formation of the Infinitive ...... 149 3.14 Formation and uses of the converbs ...... 151 3.14.1 Formation of the t-converb ...... 151 3.14.2 Formation of the m-converb ...... 152 3.14.3 m-converb vs. t-converb ...... 153 3.14.4 The converbal linker -tanə ~ -ta ...... 154 3.14.5 Functions of the converbs ...... 155 3.15 Verbal negation ...... 156 3.15.1 Negated Perfective ...... 156 3.15.2 Negated Imperfective ...... 157 3.15.3 Negated Jussive ...... 158 3.15.4 Prohibitive ...... 159 3.15.4.1 Negated Jussive vs. Prohibitive ...... 159 3.15.5 Negation of Past Imperfective ( + banə) ...... 159 3.15.6 Negation of Past Perfective ( + banə) ...... 160 3.16 Depalatalization ...... 161

iii 3.17 The verb bar ‘say’ ...... 166 3.17.1 bar ‘say’ as quotative verb ...... 166 3.17.2 bar ‘say’ as complementizer / subordinator ...... 168 3.17.3 Other subordinate clauses with bar ‘say’ ...... 170 3.17.4 Phrasal verbs with bar ‘say’ ...... 171 3.17.5 The infinitives wəβəru and bɨrotu ...... 173 3.17.6 Further idiosyncratic uses of bar ‘say’ ...... 174 3.18 Tense, aspect, and mode (TAM) ...... 174 3.18.1 Perfective ...... 174 3.18.1.1 Main verb marker -m ...... 177 3.18.2 Imperfective ...... 178 3.18.3 Jussive ...... 179 3.18.4 Analytical forms with past auxiliary banə ~-βa ...... 179 3.18.5 Future tense ...... 181 3.18.5.1 Use of the two Futures ...... 183 3.18.5.1.1 Future vs. Imperfective ...... 183 3.18.5.1.2 Definite Future vs. Indefinite Future . 184 3.19 Verbs of being ...... 188 3.19.1 ʽHaveʼ and ʽmustʼ ...... 193

4 Nominal morphology 195 4.1 Derivational nominal morphology ...... 195 4.2 Number and gender ...... 204 4.2.1 Number ...... 204 4.2.2 Associative marker nə- ...... 207 4.2.3 Gender ...... 208 4.2.3.1 Feminine derivational morpheme -wət . . . . . 210 4.3 Pronouns ...... 211 4.3.1 Independent pronouns ...... 211 4.3.2 Demonstratives ...... 212 4.3.2.1 The demonstrative element -kk- ...... 215 4.3.3 Further occurrences of the demonstrative elements . . . . 216 4.3.3.1 Presentatives ...... 216 4.3.3.2 Spatial adverbs ...... 216 4.3.3.3 Clause conjoining with xɨ and zɨ ...... 217 4.3.4 Indefinite pronouns ...... 219 4.3.5 Question words ...... 220 4.3.6 Possessives ...... 227 4.4 Definiteness ...... 233 4.4.1 Definite articles ...... 233 4.4.1.1 The definite article -we ...... 236 4.4.2 Use of articles and marking of definiteness ...... 237 4.4.2.1 Direct anaphora vs. associative anaphora . . . . 237 4.5 Numerals ...... 239

iv 4.5.1 Cardinal numerals ...... 239 4.5.2 Ordinal numerals ...... 241 4.6 Days of the week ...... 241 4.7 Nominal affixes and subordinators ...... 242 4.7.1 Affixes and nouns: prefixes ...... 243 4.7.1.1 Attributivizer yə- ...... 243 4.7.1.2 Recipients, /, objects marked by yə- . . 246 4.7.1.3 Peripheral constituents marked by yə- . . . . . 246 4.7.1.4 Locative bə- and instrumental bə- ...... 248 4.7.1.5 Comitative tə- and ablative tə- ...... 249 4.7.1.6 Combinations with postpositions ...... 250 4.7.2 Affixes and nouns: suffixes ...... 251 4.7.3 Affixes and verbs: subordinate clauses ...... 252 4.7.3.1 ʽRelative verbʼ ...... 252 4.7.3.2 Relative clauses ...... 254 4.7.3.3 Temporal clauses ...... 254 4.7.3.4 Complement clauses ...... 256 4.7.3.5 Conditional clauses ...... 257 4.7.3.6 Purposive clauses ...... 257 4.7.3.7 Causal clauses ...... 259

Acknowledgements

The present thesis has been realized as part of the project ʽFunctional Typology of Ethiopian Languagesʼ (ʽFunktionale Typologie äthiopischer Sprachenʼ) granted by the Swiss National Science Foundation (no. 100012-109306 and no. 100012- 117861). None of my work would have been possible without the generous help of so many people. I would like to express my deepest gratitude to Yunus Hassen who became my friend and most important Gumer teacher. Without his patience during the countless hours he spent with me, this grammar would not be what it is. I am also greatly thankful to my other Gumer teachers Fedlu Hassen, Hayatu Mohammed, Ruth Belachew and in particular Wondimu Aregga, who first introduce me to the Gumer villages and their people. I appreciate the hospitality of the families, relatives and friends of Wondimu, Yunus and Fedlu who opened their homes for me. Special thanks go to the many Gumer speakers who let me record stories and narrations, the most important contributors being Hassen, Musamma and Zerga. In Ethiopia, Ronny Meyer played a key role both on a personal and a linguistic level. I highly appreciate his support and I am very grateful for the accomodation he provided me with for such a long time. I also wish to express my thanks to Hans and Gabi Schaumberger for their helping hand. Very important in establishing the first connections with Gumer speakers were Girma A. Demeke, Debela Goshu, Abubakr Sherifo and Seid Ahmed. I would like to thank my colleagues at the Department of Comparative Lin- guistics (the former Department of General Linguistics) at the University of Zürich for the many linguistic discussions and their support in general: Silvia Zaugg- Coretti, Per Baumann, Alexandra Herdeg, Mathias Jenny, Rafael Suter, Tobias Weber and Fernando Zúñiga. And I am endebted to the late Karen H. Ebert who initiated the SNSF project and gave me the chance to undertake this adventure. I am also thankful for the numerous valuable discussions on various linguistic topics I had over the years with Ronny Meyer, Joachim Crass, Ian Maddieson, Stephan Schmid, Ya’ar Hever, Christian Rapold and many others, including the participants of workshops and conferences.

vii Abbreviations

* ungrammatical form ? questionable form ¿ possible but rather dispreferred form √ root 1, 2, 3 first, second, third person 1, 2, 3, (4) first, second, third, (fourth) consonant in a root Ѧ (spontaneous) Amharic loan (in glosses)  ablative tə-  focusing/coordinating ʽalsoʼ -m  Amharic  associative nə-  attribitivizer yə- . past auxiliary banə~ -βa . past  and past  banə ~ -βa  benefactive  comitative tə-  complementizer  copula  converb . m-converb . t-converb  dative yə-  definite (article)  demonstrative  directional -nyə  verb of existence f feminine  frequentative . definite future -te . indefinite future -ʃə  goal -e  imperative  instrumental bə-  imperfective  impersonal  intransitive verb  jussive  converbal linker -ta~ -tanə  locative bə- m masculine  main verb marker -m  malefactive-locative-instrumental

viii  negation  (primary) object p plural  perfective  plural  possessive  pragmatic praticles -a, -ʃ, -x  prohibitive  purposive -e  relativizer yə- s singular  subject  singular  dummy head ‘thing’ kʼar  transitive verb

1 Introduction

1.1 Scope of the thesis This thesis aims at describing the phonology and – to a greater extent – the morphology of Gumer, one of the many varieties that constitute the Gurage cluster of Ethiosemitic languages and dialects in Central Ethiopia. Gumer is closely related to Chaha, which is probably the best known and most studied among all (Western) Gurage varieties. In contrast, Gumer has virtually not received any attention so far, the main reason being that it is usually only considered a subdialect of Chaha or even the same as Chaha. Yet, despite the relatively numerous publications on Chaha, there is still no work deserving the name ‘grammar’ – let alone ‘reference grammar’ – of Chaha or any other Western Gurage variety.¹ The present study pursues two main purposes by which I hope to fill some of the existing gaps and to contribute new insights to the overall knowledge we have about Gurage languages. On the one hand, this is the first time an extensive amount of data are collected that are specifically and exclusively Gumer and not, at best, “greater Chaha”. In order to be able to establish the (presumably small) differences between the two varieties, it was important to focus on those domains of the grammar that usually reveal dialectal differences the most. For that reason, the scope of this thesis is limited to a detailed a thorough documentation and description of the phonology and morphology of Gumer. This does not exclude the possibility that there are also differences in syntax, but it is quite probable that they would be even more subtle. However, for such a study a big corpus (of spoken Gumer) would be necessary, a task that exceeds the capacity of this thesis by far. On the other hand, as the title “Towards a grammar of Gumer” suggests, this thesis also intends to be an initial step towards a first comprehensive description of a Western Gurage variety. With this in mind, this study provides also occasional functional explanations supplementing the otherwise formal orientation of this study.

1.2 Gumer people and language Gumer is spoken in the Gumer wərəda (district) in the Gurage Zone of the South- ern Nations, Nationalities, and People’s Region (SNNPR) of Ethiopia, approximately 130 km straight-line distance south-west of the country’s capital Addis Ababa. Alongside the gravel road from Welk’it’e to Hosaina, the Gumer speaking area roughly extends from Jemboro (ʒəmbʷərə) to the town of Arek’it’ (arəkʼɨtʼ). Midway between these two localities lies Mazoriya,² a place where the road turns south. In twenty minutes walking distance north of Mazoriya there is a big mar-

¹ There are two grammars of Eastern Gurage languages, namely on Zay (Meyer 2005b) and on Wolane (Meyer 2006). ² More accurately it is called Gumer Mazoriya or Bole Mazoriya, məzoriya meaning simply ‘detour’ or ‘turning point’ in Amharic.

1 ket place with supra-regional importance called Bole or ewə gəbya. The region is about 3,000 meters above sea level and rather green and fertile. The landscape is characterized by successions of elongate ridges of hills on top of which the typical round houses are built alongside a broad middle way or public space called jəffʷərə. The Gumer extensively cultivate əssət (enset, ensete ventricosum, “false banana”), the common staple crop of the Gurage and many other peoples of south- west Ethiopia (cf. Shack 1966). The self- of the Gumer [gumər] is gʷəmarə.³ Their neighbors are speakers of other Gurage varieties: Chaha (cəxa) to the north-west, Ezha (əʒa) to the north, Muher (mʷəxɨr) to the north-east, Siltʼe⁴ (sɨltʼe) to the east, Gyeto (gʸəta) to the south, and Inor (ɨnoːr) to the south-west. The exact number of Gumer speakers is not known. According to the 2007 population and housing census (FRE 2008: 76) there are about 80,000 people living in the Gumer wərəda. However, since the political borders and the linguistic confines do not necessarily correspond, the number of Gumer speakers may diverge considerably. According to the preceding population and housing census of 1994, which also has statistics of languages, there were 139,000 speakers of ‘Sebat Bet Gurage’ (of which Gumer is one variety, see section 1.3) in the Gumer wərəda. However, the Gumer wərəda seemingly has been divided recently⁵ and was bigger in 1994 including most Gyeto and parts of other groups. Lewis (2009) does not give any numbers for Gumer, but only for the other dialects of ‘Sebat Bet Gurage’ with a total of 440,000 speakers. Subtracting Lewis’s 80,000 Gyeto from 139,000 there would be roughly 60,000 Gumer speakers. This is less than the number of inhabi- tants of the Gumer wərəda in 2007, but taking into consideration rapid population growth a total of 80,000 speakers still seems quite reasonable. Nevertheless, due to the lack of reliable new statistics concerning languages, as well as other uncertain factors as for example migration to bigger cities outside the area or the fuzzy dialect borders especially between Gumer and Chaha, the actual number of speakers could differ significantly, even by several tens of thousands. The Gumer mostly adhere to the Christian or Muslim religion. The 1994 population and housing census reports that about 80% of the population in the Gumer wərəda is Muslim. This is in contrast to the neighboring Chaha wərəda, where more than 50% are Christians.

³ Often Gumer is also written Gumär, <ä> representing [ə]. Other spellings for gʷəmarə I have come across in the literature are: Gʷämära, Gʷämara, Gʷəmaro, Gomaro, Gwemarra. ⁴ The Silt’e do not consider themselves Gurage (any more). In fact they are now organized in their own political administrative zone which officially split from the Gurage zone in 2001 (Nishi 2005). Also other groups do not call themselves Gurage, but refer to themselves by their respective group name (Meyer 2011: 1221). ⁵ There are no maps or detailed figures available to me.

2 1.3 Classification of the language Gumer is a variety of Gurage and belongs to the South-Ethiosemitic (also known as South Ethiopic) division of the Semitic branch of the Afroasiatic phylum. The term Gurage does not stand for a uniform linguistic entity, but is used as a (geo- graphical) designation for the southernmost Semitic speaking peoples surrounded by Cushitic and Omotic languages in a region roughly bordered by the “Rift Valley lakes in the east, the River Awash in the north and the River Gibe in the west and southwest” (Meyer 2011: 1220, referring to Hetzron 1972: 6f.; see also Goldenberg 2005: 924).

Afro-Asiatic > Semitic > South > Ethiopian > South → South-Ethiosemitic

South-Ethiosemitic

Outer SE Transversal SE

n-group tt-group Harari-East Gurage Amharic Argobba †Gafat Dobbi Muher Western Gurage East Gurage Harari Kistane | Zay Mesqan PWG/CWG Silt’e

Peripheral Western Gurage Central Western Gurage | | Gyeto Gumer Inor Chaha Endegeñ Gura Ener Ezha †Mesmes

Figure 1: South Ethiosemitic

As can be seen in Figure 1, the Gurage varieties belong to both subbranches of South-Ethiosemitic, namely Outer South-Ethiosemitic and Transversal South- Ethiosemitic. Transversal South-Ethiosemitic consists of Amharic – the national language of Ethiopia – and the group Harari-East Gurage. East Gurage contains the languages Zay and Silt’e, including Wolane and others. The varieties of Outer South-Ethiosemitic are all considered to be Gurage except for the extinct Gafat. They share a number of typological features different from East Gurage and are called Gunnän-Gurage by Hetzron (1977). Genetically, Gafat, Kistane (=Soddo)

3 and Dobbi (=Gog(g)ot) form the n-group.⁶ The tt-group consists of Muher on the one hand and Western Gurage on the other hand. Western Gurage, in turn, contains Mesqan and the two groups Central Western Gurage and Peripheral Western Gurage. As for Gumer, it is part of Central Western Gurage together with Chaha, Ezha and Gura. The detailed classification of the Gurage varieties and in particular their status as dialects or distinct languages is still in debate. Muher, for example, has also been grouped together with Kistane and Dobbi as Northern Gurage (though on typological grounds, cf. Hetzron 1977: 22). In Lewis (2009), Muher appears at a completely different position. The tt-group is rearranged and divided into four different languages with several dialects: the extinct Mesmes, Mesqan, Inor with the dialects Endegeñ and Ener, and Sebat Bet Gurage with the dialects Chaha, Ezha, Gumer, Gura, Gyeto and Muher. The term Sebat Bet (or səβat bet) ‘Seven Houses/Tribes’, however, is usually not used as a language name but rather refers to a political confederation of the seven (main) groups of Western Gurage.⁷ Based on my own experiences with some of the Gurage varieties, I claim that Muher is the most divergent of these assumed dialects of Sebat Bet Gurage, whereas Gumer, Chaha and (to a lesser extent) Ezha are rather similar.

1.3.1 Gumer vs. Chaha There is no question that Gumer and Chaha are very closely related. Leslau (1996: 111) states that “there is no difference between Gumär and Čaha”, but this position is slightly too strict. There are some differences that indicate that these two varieties are best regarded as a dialect continuum, or following Hetzron (1977: 4f.) “Chaha and Gumer [...] should be considered two dialects of the same language”. Rose (2007: 403) mentions that Gumer and Gura are subdialects of Chaha. How- ever, according to a Gumer speaker, Gura is difficult to understand,⁸ while there seem to be no problems concerning the mutual intelligibility between Ezha and Gumer. This fact suggests that probably the whole Central Western Gurage group forms one dialect cluster. The Gumer and the Chaha are not only closely connected linguistically, but also in the Sebat Bet Gurage confederation they are represented by one common delegate and often subsumed under the term əgrangət cəxa,⁹ as expressed in the

⁶ According to Hetzron (1977: 23f.), the division of Outer SE into the n-group and the tt-group is based on the different shape (i.e. -n and -tt) of the so-called main verb markers after (historically) long vowels. Note, however, that Mesqan and Central Western Gurage (thus also Gumer) have no traces of such a morpheme. ⁷ It is not entirely clear which groups belong to Sebat Bet (cf., for example, Shack 1966: 205f., Hetzron 1977: 4ff., Leslau 1979a: ix ff.). ⁸ One reason for Gura to be difficult to understand seems to be that any pre-stress consonant may be articulated with a glottal constriction (Hetzron 1977: 43), for example Gura bʔarəcʔənʔˈa vs. Gumer barəcənam ‘she told her’. But also the Gura texts in Hetzron (1977: 216ff.) suggest that it is more deviant from Chaha than Gumer. ⁹ The word əgrangət is composed of əgr ‘foot’ and angət ‘neck’, which refer to the Chaha and

4 following Gumer sentence (1) taken from an account about the origins of the Sebat Bet Gurage.

(1) cəxa tə-gʷəmarə ɨmmat-u, əgr-angət cəxa yɨ-wr-i. Chaha COM-Gumer only.one-COP.3smS foot-neck Chaha 3S-say.IPFV.IPS-3smO ‘Chaha and Gumer are the same, one calls them Egranget Chaha.’

Despite the close relation between Gumer and Chaha, people know which group they belong to.¹⁰ Nevertheless, it seems hardly possible to draw a clear dialectal boundary on linguistic grounds between them, as the speech of some Gumer speakers closer to the Chaha area tends to exhibit less features typical of Gumer. The boundary to the Ezha in the north seems easier to identify: there is, for example, a small village called yətəjyə, where a few of the about sixteen families speak Gumer, the majority speaking Ezha.¹¹ The identified differences between Gumer and Chaha mainly concern phonology and morphology. They do not seem to diverge syntactically, but to substan- tiate this claim more research is needed. The following is a short summary of the differences (cf. Völlmin 2009 for a more detailed description). The sound [l] is rare in Gurage, but in Gumer it occurs more often than in Chaha. The enclitic present-tense copula (↗ 3.19) of 3pm and 3pf are -lo and -ləma in all environments, whereas in Chaha they are -ro/-rəma after vowels, -no/-nəma after consonants and -lo/-ləma after [r] (Leslau 1983: 14). Likewise, the benefactive suffixes (↗ 3.12) of the third persons singular and plural throughout start with an [l] (-lə/-la/-lo/-ləma) in Gumer, whereas in Chaha they have [r] instead of [l]. Similar to the copula, in Chaha there is an allomorph with [l] after final [r] only (Hetzron 1977: 38). The so-called ‘heavy’ set of the malefactive suffix (↗ 3.12) have a free variant with [k] instead of [p] in Chaha (Banksira 2000: 262). This k-variant seems to be the less frequent one in Chaha and, according to my own observations, absent in Gumer. The labialization of [β] always results in [w] in Chaha, but in Gumer also [bʷ] is heard. Chaha has the mid-open vowels [ɛ] and [ɔ]. They are usually the result of the combination a+y/i and a+w/u respectively, either within a lexeme or across morpheme boundaries. In Gumer, these combinations more often are realized as the diphthongs [ay] and [aw], but [ɛ] and [ɔ] are also heard (↗ 2.2.1). A characteristic of Ezha is its voiced geminates. In Chaha, on the other side, most geminates have undergone degemination and often also devoicing. Gumer represents a mixed type and stands between these two variants: (historical) gem-

Gumer respectively indicating their relative altitude of settling, i.e. the Gumer are above the Chaha geographically (cf. Shack & Marcos 1974: 39). ¹⁰ Note, for example, that in the mid 19th century the Chaha belonged to the confederation of Amist Bet Gurage (Five Houses), the precedessor of Sebat Bet Gurage, but not the Gumer (Shack 1966: 205). ¹¹ It is a well-known (and amusing) fact among the locals that the word jɨwə means ‘tail’ in Gumer, but ‘penis’ in Ezha. The Gumer speakers in yətəjyə try to avoid jɨwə given the presence of Ezha speakers in the same village.

5 inates have been devoiced as in Chaha, but degemination has not taken place under certain conditions. Most importantly, this concerns the penultimate radicals of verbs, which have preserved gemination whenever the final radical is short, i.e. in particular [r], or totally missing (↗ 3.3.1).

1.4 Previous studies Due to the fact that Gumer usually is considered a subdialect of Chaha and not the other way round (cf., for example, Leslau 1969: 96), it has not received much attention so far and there are only scant data available. Hetzron (1977) has one Gumer text, mentions several characteristic features of Gumer and even presents a paradigm of the present-tense copula. The same paradigm, which, however, does not conform to my data, is also found in Hetzron (1972). Meyer (2007) and Crass & Meyer (2008) contain a few Gumer data, and Meyer (2005a) is a short encyclopedia article on Gumer. Own publications on Gumer prior to the completion of this thesis are “Some dialectal differences between Gumer and Chaha (Gurage)” (Völlmin 2009), “Benefactives and malefactives in Gumer (Gurage)” (Völlmin 2010a), and “The so-called converbs in Gumer (Gurage)” (Völlmin 2010b). Chaha, on the other hand, is probably the best documented Gurage variety. An early (and for that reason completely outdated) publication presenting very few data and a vocabulary list is Chiarini (1887). Since then, Chaha has repeatedly been the subject of studies, while other varieties have only been treated punctually or not at all. Wolf Leslau, a pioneer in Gurage studies, contributed many publications. The most important contribution is an etymological dictionary of Gurage in three volumes (Leslau 1979a,b,c), containing a great amount of lexical data of twelve Gurage varieties including Chaha, but not Gumer. Other works treating specific topics in Chaha grammar are Leslau (1948, 1964, 1967, 1992, 1997); Leslau (1950, 1966, 1983) also contain texts. Further articles dealing with Chaha include Banksira (1999a,b), Ford (1991, 2003), Kenstowicz & Banksira (1999), Lowenstamm (1996, 2000), McCarthy (1983, 1986), Petros (1993, 1994, 1996a,b,c), Prunet & Pet- ros (1996), Rose (1994, 2000), and Ueno (2004). Finally, among the newest publications are Rose (2007), a concise summary of the Chaha morphology, and Banksira (2000), a noteworthy and quite exhaustive study of Chaha morphophonology.

1.5 Data and method Most of the data presented in this description were gathered during four different trips to Ethiopia between February 2006 and April 2008 with a total of six months in the field. Three subsequent trips in 2010 and 2011 of one month duration each were used to cross-check, verify and add data. In the early phase, the data col- lection took place mainly in Wolkite, the administrative center of the Gurage region, in a later phase more often in Addis Ababa, and of course in several Gumer villages in all phases. The villages – such as keβaro, cʼacʼo in the north-western part, and dəwəʃe and xəttərat in the central part of Gumer – are all located on the

6 northern side of the road Wolkite-Hosana between Jemboro (ʒəmbʷərə) and Bole (ewə gəβya). Gumer regions further south and in direction of the town of Arekʼitʼ (arəkʼɨt’) could not be included. The Gumer data were gathered mainly in two ways: on the one through elic- itation – primarily for vocabulary, word forms, verb paradigms and simple sentences –, on the other by means of recording and transcribing conversations and stories. In a few recording sessions stimuli were used such as pictures and short clips of the projects ‘/ʼ (Bohnemeyer, Bowerman & Brown 2001) and ‘ʼ (Bowerman, Gullberg, Majid & Narasimhan 2004). Example sentences taken from recordings are as close as possible to the original utterances, but note that they are often edited and normalized, eliminating unnecessary repetitions, hesi- tations and obvious (performance) mistakes as well as mending unfinished sentences with the help of language consultants. The main language consultants (the so-called ‘informants’) and contacts to the Gumer villages were all young males in their (early or late) twenties residing in Wolkite or Addis Ababa. The story- tellers of the recorded and transcribed texts were mostly older men above the age of fifty.

7 2 Phonology

2.1 Consonants Figure 2 shows the Gumer consonant inventory. Note (a) that the symbols as they are used here are based on the IPA alphabet, but with some modifications (for specifications of the phonetic qualities of some of them see below); and (b) that not all of these consonants are fully independent phonemes: some of them have a restricted distribution or can be analyzed as allophones (↗ 2.1.1-2.1.3).

labial alveolar velar stop p b t tʼ d k kʼ g palatalized c cʼ j kʸ kʼʸ gʸ labialized pʷ bʷ kʷ kʼʷ gʷ fricative f s z x palatalized ʃ ʒ xʸ labialized fʷ xʷ nasal m n labialized mʷ approximant β liquid r l glide w y

Figure 2: Gumer consonant inventory

Like many languages in the Horn of Africa, Gumer has ejectives (Cʼ) in addition to the plain consonants (C). Labials and velars can be labialized (Cʷ), whereas alveolars and velars have palatalized counterparts (palatalized velars written with a superscript ʸ, i.e. Cʸ). The consonant symbols that are not in accordance with IPA include: the palatal affricates c, cʼ and j, which correspond to [ʧ], [ʧʼ] and [ʤ]¹²; the glide y, which stands for [j], and accordingly the symbol for palatalization, which is a superscript ʸ rather than [ʲ]; and thus the palatalized velars [cʼ], [c], [ɟ], [ç], which are written here as kʼʸ, kʸ, gʸ, xʸ. The quality of r seems to come closest to an approximant [ɹ], especially in coda position (Banksira 2000: 131). Finally, according to Banksira (2000: 2) the bilabial represented as β is an approximant [β]̞ in Chaha. In Gumer, however, this consonant seems to be slightly more closed.

¹² Banksira (2000: 2) states that cʼ, c and j do not consist of two phases (i.e. stop and fricative) and differ in pronounciation from clusters tʼ+ʃ, t+ʃ and d+ʒ, respectively. For instance, the realization of tʃ in yɨ-t-ʃəkət ‘it gets better’ is different from c in yɨ-cəkkɨr ‘he cooks ()’. While I agree that their pronounciations differ (the ‘simple’ ones are ‘softer’), I think that one cannot readily compare c with t+ʃ. In the latter case the sound is made up of two morphemes, a fact that may explain its ‘stronger’ realization.

8 This finding is supported by (Hetzron 1977: 50), who mentiones that b does not change to β in Gumer. (On the distribution of β/b/p see also section 2.1.1). Additionally to the above consonants, Gumer speakers marginally also use in loan words (usually taken from Amharic) the ejectives pʼ and sʼ as in itopʼya ‘Ethiopia’¹³ and məsʼaf ‘book’, and the palatal nasal ɲ, mainly in the (Amharic) suffix -(ɲ)ɲa for language names like gʷɨragʷiɲa ‘Gurage (language)’ or amarɨɲɲa ‘Amharic’.¹⁴ Further, I am aware of one instance of a glottal stop ʔ (that is rather strong and clearly differs from a ‘normal’ onset of words beginning with vowels), namely in ʔar, which is used in variation with kʼar ‘thing’.¹⁵ The following examples provide minimal or near-minimal pairs for most of the consonants in the contrasts voiceless vs. voiced (2), plain vs. ejective (3), plain vs. palatal (4), plain vs. labialized (5), and others (6).

(2) voiceless vs. voiced tɨrəttərəm ‘he tore’ dɨrəttərəm ‘it thickened’ tən ‘smoke’ dən ‘stomach’ kɨβɨr ‘respect ()’ gɨβɨr ‘objects, goods’ kəppam ‘he folded’ gəppam ‘he entered’ cəkkərəm ‘he boiled’ jəkkʷərəm ‘it wilted’ gʷəcə ‘appointment’ gʷəjə ‘hole, ditch, pit’ ərəkʸ ‘I throw’ ərəgʸ ‘throwǃ ( 2sm)’ kʸəkkʸərəm ‘he held something in armpit’ gʸəkkʸərəm ‘he straightened out’ kʷəncɨwə ‘small jar’ gʷəncə ‘hyena’

¹³ The Gumer pronounciation of ‘Ethiopia’ (Amharic: ityoppʼɨya) varies: itopʼya ~ ɨtʼopʼya ~ ɨtopʼya. ¹⁴ Natively, sʼ is replaced with tʼ (məsʼaf ~ mətʼaf ‘book’, sʼaf ~ tʼaf ‘write’) and ɲ with n (gʷɨragʷiɲa ~ gʷɨragʷina ‘Gurage (language)’, amarɨɲɲa ~ amarina ‘Amharic’). ¹⁵ The replacement of ejectives with glottal stops (debuccalization) occurred on a more regular basis in other Gurage varieties (cf. Hetzron 1997: 537), compare for example Inor bəʔər with Gumer bəkʼɨr ‘hundred’ or Inor ãʔãr with Gumer ɨntʼar ‘stick’. The case of Gumer ʔar ~ kʼar is probably a recent phenomenon seemingly found mainly with younger speakers. According to an informant, ʔar sounds more like Chaha. The phonological reduction may be explainable to some extent with its high frequency.

9 kʷər ‘saddle’ gʷərə ‘season of agricultural activities’ sənam ‘he arrived’ zənam ‘he sowed’ sassam ‘he became thin’ zassam ‘he acted mad’ ʃɨr ‘legal argument’ ʒɨr ‘measurement of land’ baʃə ‘sickness’ bʷəʒə ‘lightning’ dappʷa ‘kind of bread made from əssət’ sambʷa ‘lung’ gʷɨppa ‘layer of stem of əssət’ abba ‘father’

(3) plain vs. ejective tɨkkʷɨr ‘cooked, boiled; a meal with cabbage and meat’ tʼɨkʼʷɨr ‘black’ attərəm ‘he spent the night’ attʼərəm ‘he made a fence’ cənəcɨm ‘she came’ cʼənəcɨm ‘she gave birth’ gʷəncə ‘hyena’ kʼʷəncʼə ‘frog’ kaʃ ‘divideǃ ( 2s)’ kʼaʃ ‘throw awayǃ ( 2s)’ fəkkam ‘he went away’ fəkkʼam ‘he split (wood)’ kʸəʃə-m ‘also a joke’ kʼʸətʼəm ‘he got tired’ akʸəsəm ‘he joked’ akkʼʸəm ‘he crunched grain’ kʷər ‘saddle’ kʼʷərə ‘roasted grain’ tɨkkʷɨr ‘cooked’ tʼɨkʼʷɨr ‘black’

10 (4) plain vs. palatal tɨtot ‘may she work’ tɨcot ‘she works’ tənəm ‘he swore (an oath)’ cənəm ‘he came’ tʼəkkʼʷərəm ‘it became black’ cʼəkʼʷrərə ‘early morning’ kʼəntʼəm ‘he punished’ kʼʷəncʼə-m ‘also a frog’ tədanəgom ‘they hit one another’ təjanəgom ‘they bumped into each other’ dəppərəm ‘he added’ jəppərəm ‘he finished’ sat ‘hour’ ʃat ‘bee-hive’ səkkərəm ‘he got drunk’ ʃəkkərəm ‘he changed’ zər ‘rainy season; seed’ ʒəra ‘light brown (cattle)’ tɨzor ‘let her go around’ tɨʒor ‘she goes around’ tafʷəkɨk ‘you (sm) crawl on all fours’ tafʷəkʸɨkʸ ‘you (s) crawl on all fours’ məkkərəm ‘he gave advise’ məkkʸərəm ‘he set fire’ tɨfəkʼ ‘she scrapes’ tɨfəkʼʸ ‘you (2s) scrape’ kʼəttʼam ‘he punished’ kʼʸətʼəm ‘he got tired’ gɨβt ‘hal’ gʸɨβat ‘horse riding’ gənəzəm ‘he grew old’ gʸənəzəm ‘he cut into big slice’ təxəttərəm ‘he covered’ təxʸəttərəm ‘he followed’ yə-x ‘of that’ yəxʸ ‘here you (s) areǃ’

11 (5) plain vs. labialized bər ‘door’ bʷəra ‘cow with white spot on forehead’ bɨʃa ‘red’ bʷɨʃe¹⁶ ‘evil’ jippə ‘mat’ cʼɨppʷə ‘parings of the əssət-root’ tʼəppam ‘he skinned’ tʼəppʷəm ‘he sucked’ kərə ‘day’ kʷər ‘saddle’ cəkkərəm ‘he boiled’ jəkkʷərəm ‘it wilted’ kʼənəm ‘he vanished’ kʼʷənəm ‘he roasted kʼəmətʼ ‘shame’ kʼʷəmət ‘calabash’ gɨppe ‘enclosure’ gʷɨppa ‘layer of stem of əssət’ gɨyə ‘dog’ gʷəyə ‘main house’ axə-m ‘also you (2sm)’ axʷəm ‘he spilled ()’ xɨmmu ‘here it is’ xʷɨmma¹⁷ ‘elbow’ cəffərəm ‘he gave a mouthful’ jəffʷərə-m ‘also public space’ tifə ‘slap into face’ tʼɨfʷə ‘bad’ mena ‘work, job’ mʷena ‘uncle (maternal side)’ yɨtəm ‘it tastes sweet’ y-ətəmʷ ‘of sister’

¹⁶ In Leslau (1979c: 161) one finds buʃe. ¹⁷ In Leslau (1979c: 365) one finds xumma for Chaha and xʷɨmma for Ezha.

12 (6) others dəm ‘blood’ dən ‘belly’ kʼʷəmmərəm ‘he became strong’ kʼʷənnərəm ‘he trimmed’ cʼamma ‘shoe’ kʼʸamma ‘guarantor’ accʼəm ‘he shut’ akkʼʸəm ‘he crunched grain’ gʷəla ‘horse pen’ gʷərə ‘season of agricultural activities’ oʃa ‘careless’ oxʸə ‘Ohye (subgroup of Gurage)’¹⁸ bɨʃa ‘red’ bɨxʸə ‘mourning, funeral’ əxʷa ‘now’ əkkʷa ‘today’ gɨyə ‘dog’ gɨwa ‘cheapʼ tɨyəm ‘he was visible’ tɨwə-m ‘also hard’

2.1.1 Distribution of β, b and p Following Banksira (2000: 7), β, b and p are largely allophones of /β/. The voiceless plosive [p] generally originates from a geminated /β/. It occurs as the mutated (↗ 3.3.1) penultimate radical of a verb root, either geminated (7a) or degeminated (7b), as the heavy malefactive (↗ 3.12) suffix (7c), but also in nominals (7d), which usually have counterparts in other Guarge varieties with bb instead of pp or p.

(7) a. səppərəm (cf. yɨsəβɨr ‘he breaks’) ‘he broke’ b. sɨrəpətəm (cf. yə-səmbɨt ‘let him spend some time’) ‘he spent some time’ c. cəno-pi-m (cf. cənə-βi-m ‘he came to my detriment’) ‘they came to my detriment’ d. gʷəppay (cf. Ezha gʷəbbe, Leslau 1979b: 90) ‘brother’

¹⁸ More precisely, the Ohye (or Oçe) are a subgroup of Muher (Meyer 2011: 1222).

13 There are occasional loans with p (especially word-initially); however, a p that does not originate in a geminate β is usually replaced either by f as in folis ‘police’ or by b as in basta ‘spaghetti’ (< pasta) and bawnd ‘ten birr’ (< pound). The voiced plosive [b] (or ‘strengthened’, cf. Banksira 2000: 159ff.) occurs in geminates (there is no [ββ]) that have not been devoiced (8a), or in (underlying) voiced ‘geminates’ that have been degeminated (8b). These two instances are relatively rare, since most ‘geminate’ /ββ/ are devoiced to [p(p)] (as seen in (7)). Further, all word-initial /β/ are realized as [b] (8a)-(8c), as well as all β following a nasal (8d).

(8) a. abba ‘father’ gʸəbbərəm ‘he paid taxes’ b. tʼəbəsəm ‘he roasted’ c. bəsər ‘meat’ banə ‘he was’ d. ambɨkʼ ‘cold, flu’ wəmbər ‘live ()’ In all other cases, /β/ is realized as the approximant [β], i.e. word-internally not following a nasal (9a) or word-finally (9d).

(9) a. əcβa ‘central pillar of the house’ b. nɨβrət ‘live, living’ c. səβat ‘seven’ d. səβ ‘person’

If any prefix is added to a word-initial b, it is realized as [β] like all word-internal, ‘non-geminated’ /β/.

(10) a. bə-βarə ‘if he said’ (cf. barəm ‘he said’) b. yə-βər ‘let him say’ (cf. bərǃ ‘sayǃ’) When labialized, /β/ usually becomes [w]. Some speakers, however, occasionally realize it as [bʷ], i.e. it is strengthened (and never *[βʷ]).

(11) yɨ-β ‘he gives’ → yɨ-w-ɨn ~yɨ-bʷ-ɨn ‘he gives him’ However, it is worth mentioning that, since the distribution of [b] and [β] is predictable, it is not necessary to distinguish them in notation.¹⁹ As far as I understand, there is only one instance in Chaha where β contrasts unpredictably with b, namely in the verb tʼəβətʼ ‘take, hold’ (Banksira 2000: 52) where the penultimate radical exceptionally does not geminate or mutate, in contrast to, for instance, tʼəbəs ‘roast’ or ʃəpət ‘choose’. Based on this one exception, the distinctive use of <β> and is justified. Nevertheless, according to my observations, with many Gumer speakers it is hardly possible to distinguish the pronunciation of the voiced

¹⁹ The distinction of [b]/[β] vs. [p] is not predictable in all cases. There are geminate penultimate radicals [bb] that are not devoiced to [pp] as the rule would suggest, for example gʸəbbər ‘pay taxes’ vs. jəppər ‘finish’. Further, [p] can occur in some loans. Thus, the use of the symbol

is required and cannot be represented by .

14 bilabial in tʼəβətʼə and tʼəbəsə, i.e. they are either tʼəβətʼə and tʼəβəsə or (rather) tʼəbətʼə and tʼəbəsə. Further, in both verbs the β/b can become bʷ when labialized, as in (11), which in Chaha would be impossible with tʼəβətʼə and the normal case with tʼəbəsə. Also in the novel yəcʼamut ʃɨka <የጫሙት ሽካ> (Gabreyesus 1960 E.C.), which is written in Chaha, one can find tʼəwəʃim <ጠወሺም> ‘one roasted’ (p. 280) instead of expected tʼəbʷəʃim, i.e. the same consonant as in tʼɨwtʼɨnɨm <ጥውጥንም> ‘take him and’ (p. 29). These facts indicate that there tends to be no substantial (or phonological) difference between these bilabials that make a distinction in notation necessary. Nevertheless, in accordance with Banksira and others, β and b are both used here until more clarity is reached regarding their phonemic status in Gumer.

2.1.2 Distribution of x and k Banksira (2000: 91ff.) discusses the distribution of x and k in Chaha and argues that they are not contrastive. He assumes a phoneme /x/, of which [x] and [k] are allophones (note that [k] is also an allophone of /g/). This analysis is supported by the fact that there are probably no other (real) minimal pairs than the examples in (12).²⁰

(12) xənam ‘he put/prohibited’ axənam ‘he shouted’ kənam ‘he ascended ()’ akənam ‘he ascended ()’

Although Banksira can explain many occurrences of k, some exceptions or prob- lematic cases remain. For instance, there is no satisfactory explanation for the alternation of x and k in ɨxa ‘water’ and tɨkə ‘child’ respectively. Ignoring such difficulties, there are two main instances where k occurs rather than x. Firstly, k is the mutated (↗ 3.3.1) or geminated form of x, i.e. as the penultimate radical in verbs x alternates with k in certain templates:

(13)  nəkəβəm ‘he found’  yɨrəxɨβ ‘he finds’  yənxəβ ‘let him find’

Also, in other contexts, there is hardly ever a geminated *xx but only kk:²¹

(14) ɨkkɨm ‘merely, just’ əkkʷa ‘today’

Across morpheme boundaries x+x do not strengthen to k or kk, but remain xx (15). The Perfective subject suffix 1s -x and the 2nd person primary object suffixes -xə,

²⁰ Banksira (2000: 108) explains the occurrence of k in (a)kəna in that it etymologically stems from the ejective [kʼ]. ²¹ One exception is the verb xʷɨrəxxʷər ‘take out earwax’ (*xʷɨrəkkʷər) with geminated but not occlusivized xx as penultimate radical.

15 -xʸ, -xu, -xəma added to it, however, fuse to (non-geminated) -kə, -kʸ, -ku, -kəma (16) (↗ 3.12.1.1).

(15) manəx-xə-m (*manək-kə-m) capture.PFV-2smS-M ‘you captured’ (16) od-kə-m < *od-x-xə-m tell.PFV-1sS.2smO-M tell.PFV-1sS-2smO-M ‘I told you’

Secondly, at least in verb stems, k occurs when followed by a fricative (f, s, z, x) (17), or by the vowel a (or ə) that is the reflex of a (laryngeal) root consonant that diachronically got lost (18). If there is no following a/ə or fricative, the verb features x rather than k (19).

(17) kəfətəm (*xəfətəm) ‘he opened’ yɨrəks (*yɨrəxs) ‘he bites’ (18) kasəm (*xasəm) ‘he paid’ yɨfəka (*yɨfəxa) ‘he escapes’ (19) xənəməm (*kənəməm) ‘he stayed a year’ yɨβəxɨr (*yɨβəkɨr) ‘he loses’

If one adopts the idea that all surfacing k in verbs are underlying x, this rule explains the occurrence of k as a penultimate verb radical where according to the template the non-mutated form (i.e. x) is expected. The three forms in column (a) of (20) show the ‘normal’ pattern of a type A verb with the mutated penultimate radical only in the Perfective. The verb in column (b) is also of type A; nevertheless, it differs from the first one in that there is a fricative s following the velar, which consequently is realized as k in every case. Example (21) illustrates the same phenomenon with two verbs that have a vowel a instead of a consonantal radical. Again, both verbs belong to the same verb type A with mutation of the penultimate radical only in the Perfective, as is evident in (a) where the Imperfec- tive and Jussive feature the underlying, non-mutated g. In turn, the verb in (b) has an underlying penultimate radical x, which is realized as k in all forms, contrary to the prediction of the template of verb type A.

(20) a. b.  nəkəβəm ‘he found’ nəkəsəm ‘he bit’  yɨrəxɨβ ‘he finds’ yɨrəks ‘he bites’  yənxəβ ‘let him find’ yənkɨs ‘let him bite’

(21) a. b.  wəkkam ‘he stabbed’ fəkkam ‘he escaped’  yɨwəga ‘he stabs’ yɨfəka ‘he escapes’  yəga ‘let him stab’ yəfka ‘let him escape’

16 However, even though the occurrence of k vs. x in verbs can be predicted for almost all instances, I do not assume that the roots of verbs like nəkəs ‘bite’ and fəkka ‘escape’ with penultimate k, as well as verbs with initial k like kəfət ‘open’, contain an underlying radical x. Rather, their respective roots are √rks, √fkA and √kft, mainly due to the fact that the velar k never changes to or contrasts with x.

2.1.3 Distribution of r and n (and l) Banksira (2000: 125ff.) discusses the status of r and n (in Chaha) and shows that the contrast between them is neutralized in several cases. In particular, their distribution is complementary in verb bases, where, following Banksira’s analysis, r and n are represented by a single phoneme /r/, which is either realized as r or undergoes nasalization, i.e. it is realized as n. Nasalization in verb bases occurs in three instances: word-initially, as ‘geminate’ (or mutated, see section 3.3.1) penultimate radical, and in penultimate coda position. In parts of speech other than verb bases, however, one can find minimal pairs in non-initial position, showing the (partial) phonemic status of r and n:

(22) kərə ‘day’ kənə ‘right’ mʷar ‘share’ mʷan ‘who’

2.1.3.1 Initial nasalization Word-initially, /r/ is realized as n without exception, i.e. not only in verb bases but in all word classes (apart from occasional Amharic loans). This (diachronically emerged development) can be illustrated by the following three Gumer lexemes, which all begin in n, but have counterparts in Amharic with either initial n or r. This also includes l, which is rare in Gurage.

(23) Gumer Amharic nədədəm nəddədə ‘it burnt’ nəddam rədda ‘he helped’ naxəm lakə ‘he sent’

When preceded by any affix, an otherwise word-initial n surfaces as r, illustrated in (24) by the attributivizer yə- prefixed to a noun.

(24) nɨgʷɨs ‘king’ yə-rgʷɨs ‘of the king’

The same contrast of initial n vs. non-initial r in verb bases is exemplified in (25) with a plain Perfective verb and two derived stems. In (26), forms of the same verb with the prefixes bə- and yə- are shown, and (27) illustrates a Perfective verb in contrast to the Imperfective and Jussive, which have prefixed subject markers.

17 (25) nəgədom ‘they touched’ tə-ragədom ‘they touched each other’ at-ragədom ‘they caused to touch each other’ (26) bə-rəgʷəj-ɨn ‘if one touches me’ yə-rəgədə-ndə ‘who touched us’ (27) nəzəzəm ‘he dreamt’ yɨ-rəzɨz ‘he dreams’ yə-rzəz ‘let him dream’

2.1.3.2 ‘Geminate’ nasalization In verbs, the penultimate radical is mutated in certain templates (↗ 3.3.1, 3.4.1). The mutated form of r is n, i.e. it is nasalized. This mutation goes back to gemination (hence Banksira’s designation ‘geminate nasalization’), which in Gumer has been degeminated in most but not all instances, leaving behind a simplex n or a still geminated nn. As exemplified in (28) with the verb tənəf ‘remain’, which is based on the root √trf, the penultimate r mutates/nasalizes to n in the Perfec- tive, whereas the Imperfective and Jussive templates for this verb do not require mutation and the underlying r surfaces in these two forms.

(28)  tənəfəm ‘he remained’  yɨtərf ‘he remains’  yətrəf ‘let him remain’

2.1.3.3 Nasalization in penultimate coda position Another instance of nasalization of r to n (which cannot be accounted for by initial or ‘geminate’ nasalization) is in the coda position of the penultimate syllable within a verb base. This is illustrated with the triradical verb nətʼəkʼ ‘snatch’ (√rtʼkʼ) in (29): as expected, a word-initial n in the Perfective alternates with a word-internal r in the Imperfective. In the Jussive, however, the consonant in question is realized as n despite its non-initial position. Due to the pattern of the Jussive, the r forms the coda of the penultimate syllable within the verb base (yən.tʼɨkʼ), where it is nasalized. Note that this is in contrast to the nominal yə- r.gʷɨs ‘of the king’ in (24) above, where the r does not nasalize despite being in the same syllable position.²²

(29)  nə.tʼə.kʼəm ‘he snatched’  yɨ.rə.tʼɨkʼ ‘he sntaches’  yən.tʼɨkʼ ‘let him snatch’

²² Reportedly speakers in the southern part of Gumer close to the town of Arek’it’ rather do not apply the ‘penultimate coda position’ rule but would say, for example, yərfɨg instead of yənfɨg ‘let him be greedy’.

18 Also in quadriradical verbs like dɨrəkkər (√drgr) ‘throw away’, r nasalizes to n in penultimate coda position.

(30)  dɨ.rək.kə.rəm ‘he threw away’  yɨd.rək.kɨr ‘he throws away’  yə.dən.gɨr ‘let him throw away’ There are exceptions to the penultimate coda position rule. If the root contains the labial nasal m, the r in penultimate coda position does not nasalize:

(31) yər.məd (*yənməd) ‘let him love’ yə.gər.dɨm (*yəgəndɨm) ‘let him break something in two’ Also, contradictory to Banksira (2000: 139), I have recorded instances of verbs with two identical final radicals like nədəd ‘burn ()’ (√rdd) where the first radical r is not nasalized. However, it seems that both variants coexist in Gumer.

(32) yər.dəd ~ yən.dəd ‘let it burn’ ? yər.zəz ~ ( yən.zəz) ‘let him dream’

Further, quadriradical verbs with two identical final radicals like dɨrəzəz ‘be blunt’ (√drzz) or total reduplication like xʷɨrəxxʷər ‘take out earwax’ (√xʷrxʷr) usually do not nasalize (cf. Banksira 2000: 139f.).²³

(33) yə.dər.zɨz (*yə.dən.zɨz) ‘let it be blunt’ yə.xʷər.xʷɨr (*yə.xʷən.xʷɨr) ‘let him take out earwax’

Some verbs feature nasalized r→n even though it is not in the coda position of the penultimate syllable. Rather it forms the second but last consonant of the verb base (34). Following Banksira’s (2000: 135f.) analysis, this is the case with palatalized and labialized final consonants, which are complex, consisting of a plosive and an underlying radical y (=/I/) or w (=/U/). Thus, these two radicals together correspond to the ultimate syllable, the nasalized r originally occupying the penultimate coda position (for example *yə.zən.gi > yəzəngʸ).

(34) yəncʼ (*yərcʼ) ‘let him pluck’ (cf. nəccʼəm, yɨrəcʼ) yəzəngʸ (*yəzərgʸ) ‘let him speak’ (cf. zɨrəkkʸəm, yɨzrəkʸ) yənkʼʷ (*yərkʼʷ) ‘let him shout’ (cf. nəkkʼʷəm, yɨrəkʼʷ) Affixes do not affect penultimate coda nasalization: if a suffix changes the syllable structure such that an r of the verbal base becomes the penultimate syllable coda of the word (35), it does not nasalize (a). Similarly, if n is not any longer the coda of the penultimate syllable of the whole word due to a suffix, it still remains n and does not change back to r (b).

²³ One exception is dɨrəttər ‘be stout, thick’ (√drdr) with the Jussive yə.dən.dɨr (*yə.dər.dɨr).

19 (35) a. yɨ.tərf → yɨ.tər.f-o (*yɨ.tən.f-o) ‘they remain’ b. yən.tʼɨ.kʼ → yən.tʼɨ.kʼ-ə.ma (*yər.tʼɨ.kʼ-ə.ma) ‘let them () snatch’ In many cases, penultimate coda nasalization also seems to apply for nouns:

(36) a.fin.jə ‘chili’ kɨn.cɨf ‘beard’ mən.kəs ‘stomach-ache’ (cf. nəkəsəm, yɨrəks, yənkəs ‘bite’)

At least for some possible counterexamples to this rule like fɨrtʼət ‘headache’ the explanation is found in the fact that they consist of a root/base (fɨrtʼ-; cf. fəntʼ √frtʼ ‘have headache’) and a suffix (-ət) (↗ 4.1, (618)). Since suffixes do not affect the rule, the consonant in question is not in penultimate coda position and does therefore not nasalize.

2.1.4 Assimilation 2.1.4.1 Assimilation of n If followed by a consonant, the nasal n assimilates to its point of articulation, thus the symbol n as it is used here in the ‘orthography’ can stand for [n, ɲ, ŋ].²⁴

(37) alveolar sənda [sənda] ‘knife’ palatal dengʸa [deɲɟa] ‘children’ velar ɨnkʼʷɨra [ɨŋkʼʷɨra] ‘egg’

A nasal followed by a bilabial consonant becomes [m]. In addition, β after nasal always strengthens to b. Before f the nasal is pronounced [ɱ]. Both instances are written here as m.

(38) bilabial yə-mbər ‘let him live’ (cf. yɨ-rəβɨr ‘he lives’) sambʷa ‘lung’ labio-dental ɨmfas [ɨɱfas] ‘wind, air’ (cf. Amharic nəfas)

Note that the mutated penultimate radical r→n that underwent ‘geminate’ nasalization in the verb cʼənəf ‘cut off the top’ (√tʼr) seemingly does not assimilate to the directly following bilabial f in the Imperfective and Jussive.

(39)  yɨcʼənf (*yɨcʼəmf ) ‘he cuts off the top’  yətʼənf (*yətʼəmf ) ‘let him cut off the top’

²⁴ Banksira (2000: 1, 22) also uses [ɳ] classifying it as alveo-palatal. However, IPA [ɳ] stands for a retroflex nasal. Gumer does not have retroflex sounds, neither nasals nor obstruents a nasal could assimilate to. I assume Banksira’s use of [ɳ] is to signify a sound that is more palatal than [n] but less than [ɲ] to differenciate, for example, gʷəncə [gʷəɳʧə] ‘hyena’ from kʼʸɨnkʼʸə [cʼɨɲcʼə] ‘stone (of fruit), seed’. In my opinion, however, the former should be [gʷəɲʧə].

20 If n meets w (across morpheme boundaries), as is for instance the case with the negation marker an- (↗ 3.15.1) or the 1s marker -n- after any other prefix (↗ 3.11.6.3) and a verb with initial w, there are two possible outputs due to the fact that w is characterized as a labio-velar approximant with two points of articulation (Inter- national Phonetic Assiociation 1999: ix, 17). Thus, either the nasal is pronounced as a velar nasal [ŋ], or it becomes labial and the following w is occlusivized to bʷ (40).

(40) /t-n-wər/ → tɨ-n-wər [tɨŋwər] ~ tɨ-m-bʷər [tɨmbʷər] TEMP-1sS-spend.the.day.IPFV ‘when I spend the day’

Consider also the following examples, illustrating the two variants with a negated Perfective (41) and a negated Imperfective 1p (42).

(41) /an-wəttʼa-xʷ/ → [ãŋwəttʼaxʷ] ~ [ambʷəttʼaxʷ] NEG-go.down.PFV-1sS ‘I did not go down’ (42) /a-n-wəga-nə/ → [aŋwəganə] ~ [ambʷganə] NEG-1S-fight.IPFV-1pS ‘we do not fight’

Finally, (43) shows that also a β that is labialized to w due to the 3sm suffix -n produces both outputs when preceded by a nasal as the 1s prefix n- 1s.

(43) ə-βɨr ‘I say’ 1sS-say.IPFV ↓ ə-wɨn-n ‘I say to him’ 1sS-say.IPFV-3smO ↓ /b-n-βʷɨr-n/ ‘when I say to him’ TEMP-1sS-say.IPFV-3smO ↓ [bɨmbʷɨnn] ~ [bɨŋwɨnn]

2.1.4.2 Assimilation of r to n and l (across morpheme boundaries) A stem-final r and a following suffix beginning with n or l assimilate to the pseudo-geminates nn and ll respectively.

(44) jəppən-nə-m (< jəppər-nə-m) finish.PFV-1pS-M ‘we finished’

21 yɨ-xən-ni (< yɨ-xər-ni) 3smS-become.IPFV-BEN.1s ‘it is possible for me, I can’

yɨ-xəl-la (< yɨ-xər-la) 3smS-become.IPFV-BEN.3sf ‘it is possible for her, she can’

wəxe kʼal-lo (< kʼar-lo) good THING-COP.3pmS ‘they are good (things)’

fʷɨl-lo (< fʷɨr-lo)²⁵ mouse-COP.3pmS ‘they are mice’

A final l (as can occur in Amharic loans) and a following n also assimilate to the pseudo-geminate nn.

(45) an-dəwwən-nɨkʸ (< an-dəwwəl-nɨkʸ) NEG-phoneѦ.PFV-1sS.BEN.2sf ‘I did not call you (s)’

A base initial r and a directly preceding n also assimilate to nn, as the 1p prefix n- in (46) or the prohibitive marker ɨn- in (47).

(46) tɨ-n-nəβɨn-nə (< tɨ-n-rəβɨr-nə) TEMP-1pS-live.IPFV-1pS ‘when we were living’ (47) ɨn-nəgəd-xə (< ɨn-rəgəd-xə) PROHIB-touch.PFV-2sS ‘don’t touchǃ’

Assimilation of r and n to nn is even attested across word boundaries.

(48) mɨn=n-ami (< mɨr n-ami) what=1sS-do.JUS ‘what should I do’

²⁵ In Völlmin (2009: 85) there is an example xɨno ɨmar-lo ‘they are donkeys’ without assimilation. However, it seems that this only happens with slow or careful pronunciation, and if then only with nouns. Normally, final r of any word class assimilate to a following l; further examples are accʼɨl-lo (< accʼɨr-lo) ‘they are short’, bə-βəl-lo (< bər-lo) ‘they are in the river’, assɨl-lo (< assɨr-lo) ‘they are ten’, bə-βet fʷəl-lo (< fʷər-lo) ‘they are on top of the house’.

22 2.1.4.3 Other cases of assimilation Marginally, r can assimilate to s as in ɨrsɨyə ‘little’, which occasionally is pronounced as ɨssɨyə. Further, in the Jussive base of the verb tərəssa, r and s are con- tracted to təssa (tərəsa → tərsa → təssa). The assimilation of t(ə)- to the following consonant is described in section 3.6.1.2.

2.2 Vowels The Gumer vowel inventory illustrated in Figure 3 comprises the seven vowels that occur in all Gurage dialects (Leslau 1979c: xv), and very marginally the open- mid vowels ɛ and ɔ. There is no phonemic distinction of vowel length²⁶, and Gumer does not have tones.

front central back close i u ɨ close-mid e o ə open-mid (ɛ) (ɔ) open a

Figure 3: Gumer vowel inventory

In addition, vowels can also occur nasalized. Nasalization is not distinctive, but there are a few lexemes with nasal vowels as for example ãfʷ ‘bird; mouth’.²⁷ Otherwise, a vowel can become nasalized when it is followed by /n/+C, with or without complete disappearance of the nasal consonant (cf., for example, 3.15.2). This can be within a lexeme (49a), or across morpheme boundaries (49b). Rather frequent is the nasalization with the negation an- (49c). The morpheme -m of the m-converb (↗ 3.14.2) often assimilates to a following -ta(nə) becoming n, which then can nasalize the preceding vowel (49d). An instance of nasalization without complete loss of the nasal consonant is illustrated in (49e) with the benefactive marker -n followed by the object suffix -ku ‘you (pm)’.

(49) a. gɨ̃nzɨr~ gɨnzɨr ‘breakfast’ b. gə̃llo~ gənlo (< gən-l-o) ‘they are (from a) country’ c. ãxərə (< an-xər-ə) ‘it is not’

²⁶ Leslau (1979c: xix) mentions two instances of a long ɛː, bɛː bar ‘bleat’ and mɛː bar ‘bleat’, but these are clearly of onomatopoetic origin; further a long oː in oː ~ yoː ‘yes’ (cf. also Leslau 1979c: 1). For Gumer I have recorded oː and oːk, both ‘yes’, and yoː ‘yesǃ (as an answer when being called)’. ²⁷ Other examples are: ãfʷar ‘kind of tree’, ãfʷɨna ‘nose’, ãfʷat ‘odor’, ãfʷɨraxʷə ‘pregnant (of animals)’, ɨ̃fʷɨyə ‘content of both hands with palms up’, zãfʷa ‘afterbirth of cattle’. Note that they involve a labialized f. The tree name ãfʷar could be derived from of ‘mouth’, as well as ãfʷɨna ‘nose’ and ãfʷat ‘odor’.

23 d. dəwwəlõta ‘they called and’ (< dəwwəl-o-n-ta < dəwwəl-o-m-ta) e. yɨsrəβõnku (< yɨ-srə-βo-n-ku) ‘they buy for you’

An important observation concerns the frequency of the vowels. Despite the fact that Gumer exhibits a symmetrical distribution of the vowels in terms of their place of articulation, the actual frequency of the front and back vowels is significantly lower than that of the central ones. For Chaha, Banksira (2000: xxx) calculated a 1 :26 :1.23 ratio of the front, central and back vowels, respectively, i.e. the central vowels occur over ten times more often than the front and back vowels taken together. A quick calculation of my own Gumer data (both in dictionary and texts) showed a similar though slightly lower ratio, the frequency of the three central vowels ɨ, ə and a being approximately ten times higher. Contrary to Banksira, however, in my sample the front vowels occur more often than the back vowels. The reason for this disproportionate frequency lies in the special behavior of the front and particularly the back vowels. In many cases, the back vowels u and o cause labialization of a labializable consonant, and similarly the front vowels i and e can cause palatalization of a palatalizable consonant, the consequence of which is the “disappearance” of the vowels. To explain this phenomenon, Banksira (2000: xxx, 3) assumes two abstract phonemes /U/ and /I/. On the one hand, they represent the glides w and y, respectively, and on the other hand, the vowels u and i. Further, the mid vowels are in most cases biphonemic, i.e. the vowel e is a combination of /ə/+/I/, the vowel o of /ə/+/U/, and the open-mid vowels ɛ and ɔ are the fusion of /a/+/I/ and /a/+/U/. The latter two are only marginal in Gumer, where in fact this combination is normally realized as diphthongs ay and aw (↗ 2.2.1). In Banksira’s words, the explanation for the rare occurrence of the front and back vowels is then found in the fact that “the terminal features of /U/ (a phonemic element found in all back vowels) and /I/ (a phonemic element found in all front vowels) always abandon their articulators and float to dock on preceding targets” and “that they disappear and leave a trace on surrounding consonants in the form of a secondary labial or palatal articulation”. In the case of back vowels, such a ‘decomposition’ of u and o and the conse- quent labialization of a consonant can readily be illustrated with words that also exist in Amharic (loans and cognates). Thus we find for example amus ‘Thursday’, which is pronounced amʷɨs in Gumer, or the Perfective base of ‘die’, which is mot in Amharic and mʷət in Gumer. This process can be observed in the majority of words with back vowels and a preceding labializable consonant, even if there is an intervening non-labializable consonant as is the case with Amharic bəkʼlo ‘mule’ and Gumer bəkʼʷrə. Another example demonstrating the fact that the de- tachment and floating of the feature [round] is a quite natural process in Gumer is the loan word bʷəʒə ‘lightning; (name of a deity)’ which in its original Omotic

24 or Cushitic²⁸ language(s) was bazō ‘God-sky’ (Leslau 1979c: 169). Occasionally, the same process can be observed across morpheme boundaries. The pm subject marker -o is sometimes decomposed when preceded by a labializable consonant, for example yɨ-twakk-o-βa ‘they used to fight’ → yɨtwakkʷəβa. Further, in words where the preceding consonant is not labializable, labialization can also affect the following consonant as in the loans bɨrtukan ‘orange’, ʃufer ‘driver’, ʃukka ‘fork’ and mənokse ‘monk’, the Gumer pronunciation of which is bɨrtɨkʷan, ʃɨfʷer, ʃɨkkʷa and mənəkʷse~ mələkʷse. The potential of labialization to float around in words can be seen in the word meaning ‘lady’ which has two labializable consonants and two possible outputs, i.e. ɨmmʷəβecə ~ɨmməwecə. In words that do not have labializable consonants at all, labialization cannot take place, for example adot ‘mother’ or tʼona ‘strength’. Nevertheless, it is important to note that there are also instances of back vowels that do not trigger labialization of surrounding consonants, even if it were possible, as in goga ‘skin’ or kʼokʼ ‘partridge’, including some minimal pairs like bora ‘ox’ vs. bʷəra ‘cow with white spot on forehead’ or the subject marker 1s and 2pm of the Perfective, as for example in cot-xʷ-ɨm ‘I worked’ vs. cot-xu-m ‘you (pm) worked’. Consequently, the vowels u and o have to be considered phonemic (cf. also Banksira 2000: 3). As for the front vowels, the situation is not exactly the same in that there are probably no lexemes with a vowel e or i in Amharic, the equivalent of which would be decomposed resulting in a palatalized consonant in Gumer: the loan gazetʼa ‘newspaper’, for example, is still pronounced the same in Gumer and does not change to *gaʒətʼa. The element /I/ that causes palatalization as proposed by Banksira mainly represents a historically traceable (or assumed) i/y. Palataliza- tion is usually the result of diachronic developments and concerns processes on a morphophonological level rather than being a ‘natural’ synchronic process. Thus, palatalized consonants can for instance trace back to a radical y in a verb root, as kʸ in bəkkʸə ‘cry’, which derives from the common Semitic root √bky (Leslau 1979c: xxxviii) (↗ 3.4.2), or to the feminine marker of the second person singular (↗ 3.11.4), which is still an overt -i in, for example, Arabic. Nevertheless, the low frequency of the front vowels has to be accounted for in such diachronic changes. Still, there is a number of occurrences of e and i where they do not trigger palatalization, even if it were possible, as in genzo ‘ax’, wəsifə ‘awl’, or Banksira’s (2000: 3) examples yakʼetʼ ‘he trades’ and yakʼitʼ ‘let him trade’. Consequently, they have to be considered phonemic. Due to the uneven distribution of the vowels, it is difficult to find pure minimal pairs. The following list shows some vowel contrasts in near-minimal pairs or at least in similar contexts:

²⁸ Leslau (1979c: 169) states that it is “from Cushitic: Zaisse”. On Ethnologue (Lewis 2009) Zaisse is now called Zaysete (and/or is a dialect thereo) and belongs to the Omotic languages.

25 (50) sɨn ‘tooth’ səstə ‘day after tomorrow’ sin ‘cup’ sost ‘three’

sɨsə ‘thin’ neβa ‘thie’ sasa ‘thirty’ naβa ‘waist’

gɨnzɨr ‘breakfast’ gunje ‘fog’ gənə ‘land’ gonə ‘backyard near house’ genzo ‘ax’ gamʷə ‘time’

2.2.1 The open-mid vowels ɛ and ɔ As mentioned above, the mid-open vowels ɛ and ɔ occur only very marginally in Gumer. Usually, the combinations of a+y/i and a+w/u result in the diphthongs ay and aw, respectively, but some speakers, especially those closer to the Chaha area, realize them as ɛ and ɔ. This can be within a lexeme (51) or across morpheme boundaries (52b).

(51) əray (~ərɛ) ‘cows’ tʼay (~tʼɛ) ‘sheep’ wəray (~wərɛ) ‘sleep ()’ bay (~bɛ) ‘no’ təkʼaw (~təkʼɔ) ‘drink (coffee)!’ (52) a. yɨkəftəmay (~yɨkəftəmɛ) < yɨ-kəft-əma-i ‘they () open it’ 3S-open.IPFV-pfS-3smO b. kʼʸɨcʼaw (~kʼʸɨcʼɔ) < kʼʸɨcʼa-u ‘it is law’ law-COP.3smS

Note that one can also find speakers further away from Chaha who choose not to pronounce the diphthong version, but then the vowels tend to be realized more closed as e and o. In Chaha the mid-open vowels occur regularly in these contexts, and the distinction ay/aw vs. ɛ/ɔ can in principle be regarded as a dialectal difference between Gumer and Chaha. However, the fact that the mid-open vowels instead of the diphthongs were mainly recorded in Gumer speakers close to the Chaha area shows that there is no clear-cut dialect boundary. Note that all the above examples concern diphthongs in word-final position. The combination of the negation prefix a- and the subject prefix y-, for instance, is never realized as a diphthong ay but only e (only occasionally as ɛ) both in Gumer and Chaha:

(53) erəxɨβ < *a-y-rəxɨβ ‘he does not find’ NEG-3smS-find.IPFV

26 2.2.2 The central vowels ɨ and ə The two central vowels ɨ and ə need some clarification. In the ethiopistic tradi- tion (as founded by Leslau), they are represented differently: the vowel ɨ here corresponds to the ethiopistic symbol <ə> and the vowel ə to <ä>, i.e. in particular the symbol called schwa (ə) does not stand for the same vowel in these two notations. Leslau’s ethiopistic <ə> is justified for the one fact that this vowel is almost exclusively epenthetic (↗ 2.3.2), the term schwa originally referring to the epenthetic vowel (i.e. the non-underlying vowel that alternates with zero) in Tiberian Hebrew (van Oostendorp 1999). Its phonetic nature, however, is not [ə] but closer to [ɨ]. In turn, the vowel represented by <ä> is phonetically close to a prototypical [ə], which is defined by the formants f1=500 Hz and f2=1500 Hz (Ladefoged 1996: 121ff.).²⁹ In fact, my own measurements of a small Gumer sample of several speakers showed an average frequency of f1=589 and f2=1713 for the vowel ə, indicating that it is slightly lower and more fronted than an ideal [ə]. Amharic vowels do not seem to differ considerably from Gumer, i.e. percep- tually they are very similar. According to measurements by Messele (2007: 23), the Amharic lower central vowel <ä> has the formants f1=466 Hz and f2=1502 Hz, i.e. it is very close to a prototypical [ə]. These facts are in favor of the use of the (IPA) sign ə to represent this vowel both for Amharic and Gumer. The high central vowel ɨ is clearly higher than ə, but note that it is not as high as the IPA sign [ɨ] might suggest: a Russian <ы> [ɨ], for example, seems more closed. Nev- ertheless, even though the Gumer [ɨ] is lower (and probably also more fronted) than a prototypical central high vowel, its quality is still closest to [ɨ]. For all the reasons above, and also in accordance with the notation of Banksira and Rose, I have chosen the IPA symbols ɨ and ə over <ə> and <ä>.

2.2.2.1 Realizations of ɨ and ə in combination with the glides y and w The glides y and w can have an influence on the two central vowels ɨ and ə. The following examples illustrate the most important changes. The sequence əyə tends to be realized as ɛyə or ɛə. The first ə is usually fronted to an open [ɛ]. The second ə can sound rather open too, but to a lesser extent. The glide between the two vowels is very weak.

(54) asəyə → [asɛyə ~asɛə (~asɛyɛ)] ‘sell!’ zəyə → [zɛyə ~zɛə (~zɛyɛ)] ‘young woman’

Across morpheme boundaries, the sequence əwə usually becomes o (occasionally perhaps with a slightly longer realization time).

²⁹ The term ‘schwa’ is often also used to refer to the neutral, central or reduced vowel of a language in general (as opposed to the ‘phonetical schwa’). It is well known that across languages such neutral vowels can have a great varietiy of realization differing from f1=500 Hz and f2=1500 Hz (cf. Gósy 2004: 15f.).

27 (55) bə-wəxe kʼar → [boxe kʼar] ‘good bye’ yə-wəfer → [yofer] ‘of the young bull’

However, this is not the case when w is a labialized β.

(56) /nərə-βʷə/ → nərəwə (*nəro) ‘he must’ /yə-βʷər-i/ → yəwəri ~yəbʷəri (*yori) ‘one should say’

A word-initial w rounds a following ə to (approximately) [ɵ ~ ɞ ~ ə̹] and a following ɨ to [ʉ].

(57) wənad → [wɵnad ~wɞnad ~wənad̹ ] ‘mare’ wɨssa → [wʉssa] ‘bread made from əssət’

The combination of ɨ+y at the syllable coda is realized as [i]. In particular, this is the case when the Imperfective third person subject prefix y- is preceded by for example the prefix tɨ-.

(58) tɨ-y-βəra → [tiβəra] ‘when he eats’

The combination of ɨ+w at the syllable coda is approximately realized as [u], yɨwri being an example that is very frequent.

(59) yɨ-wr-i → ~ [yuri] ‘one says’

However, the pronunciation of y-uj-i ‘one tells’ (with a base initial u), for example, slightly differs from yɨ-wr-i (with a base initial w). Therefore the latter is not written as *yuri here, but yɨwri. This is in contrast to the case of ɨ+y above in (58) where the pronunciation clearly is [i] and a word like /tɨ-y-βəra/ is always written as tiβəra here.

2.3 Syllable structure and epenthesis 2.3.1 Syllable structure The Gumer syllable structure is not particularly complex. Since there are no (distinctive) long vowels, all syllables are based on a short vowel V as nucleus.³⁰ Figure 4 shows all possible patterns, but note that the ones in brackets are marginal. The common syllables are CV and CVC in all positions of a word (67)-(72). Fur- ther, there are syllables without consonantal onset, namely V, VC and VCC (60)- (66), but word-medially and word-finally they can only come about with some additional (mainly possessive) suffixes after o and e (↗ 2.3.3). The most complex syllable is CVCC (73), which, however, only occurs word-finally and only in certain constellations with respect to the sonority of the final two consonants (see

³⁰ As mentioned in the first footnote of section 2.2, there are also long vowels in three words for ‘yes’: oo (VV), ook (VVC) and yoo (CVV).

28 word-initial word-medial word-final #V (V) (V#) #VC (VC) ((VC#)) — — ((VCC#)) #CV CV CV# #CVC CVC CVC# — — CVCC#

Figure 4: Syllable structures examples (75)-(77) below). They often involve the liquid r and/or the plosive t (which is analyzed as the default segment by Banksira 2000: 9f.). Also note that, while most examples consist of one lexeme, many instances of CVCC in (73) are words ending in CVC plus an additional morpheme consisting of one consonant, or with a doubled final consonant. Other final CC clusters as well as initial CC clusters and all clusters of more than two consonants (CCC or CCCC) have to be split by means of the epenthetic vowel ɨ. The rules of epenthesis are discussed in detail in section 2.3.2. The following lists present all possible syllables, first citing three example words with the epenthetic vowel ɨ followed by one instance for the remaining (six) vowels each if available.³¹ Finally, note that the words used in this list mainly represent terminology belonging to domains of the traditional way of living such as farming, housing, tools, food etc.

(60) Word-initial syllable #V ɨ.mar ‘donkey’ ɨ.mɨr ‘stone’ ɨ.xa ‘water’ a.ka.fa ‘shovel’ e.wə ‘salty earth eaten by cattle’ ə.xɨr ‘barley, cereal’ i.wɨn ‘I give him’ o.zat³² ‘kind of porridge’ u.dɨn ‘I tell him’

(61) Word-medial syllable V (marginal) sɨ.ne.a.xu ‘your (pm) wheat’ as.so.ə.na ‘my salt’

³¹ Syllables not listed are either not existing or no example could be found. Note that some instances represent marginal cases (especially syllables with initial V), while others are abundant. ³² According to Leslau (1979c: 115) a loan from a Cushitic language.

29 (62) Word-final syllable V# (very marginal)³³ to.to.a ‘why don’t you (pm) workǃ’

(63) Word-initial syllable #VC ɨn.kʼʸər.kʼʸɨ.na ‘central rolled-up part of middle leaf of əssət’ ɨn.nə ‘a bat-like piece of wood’ ɨʃ.ta ‘women’ an.ka.se ‘a curved-handled staf’ ãfʷ.ra.xʷə ‘pregnant (animals)’ en.wa ‘dry trunk of əssət’ əc.βa ‘central pillar of the house’ iβ.xə ‘I give you (sm)’ ɨ̃fʷ.yə ‘contents of both hands with palms up’ od.xʸɨm ‘you (s) told’ ud.xə ‘I tell you (sm)’

(64) Word-medial syllable VC (marginal) kʷɨ.to.ax.ma ‘your (p) squirrel’ kʷə.to.əxʷ.na ‘their (m) mountain’

(65) Word-final syllable VC# (very marginal) bə.kʼʷəl.lo.axʸ ‘your (s) maize’

(66) Word-final syllable VCC# (very marginal) ɨnk ‘yes’ as.so.axʸʃ ‘what about your (s) salt’

(67) Word-initial syllable #CV mʷɨ.rə.ya ‘knife used for cutting leaves of əssət’ ʃɨ.ra ‘blossom of əssət’ bɨ.kʼɨl.le³⁴ ‘dried sheep leather’ ʃa.me.ta ‘slightly fermented beer made of grain’ zã.fʷa ‘afterbirth of cattle’ se.ra ‘big clay plate’ tə.βə.cə ‘knife for pulling out root of əssət’ bi.tʼɨr ‘small clay plate’ go.nə ‘area behind the house (where əssət is planted)’ tʼu.ri ‘good working, expert’

³³ This syllable structure is very marginal, the pragmatic particle -a presumably representing the only instance. Furthermore, -a could possibly also be analysed as -ʔa (CV). ³⁴ According to Leslau (1979c: 148) bɨkʼɨlle originates from a Cushitic language and describes a ‘hide

30 (68) Word-medial syllable CV ʒɨ.βan.gʸɨ.βa ‘instrument used to pound the root of əssət’ ɨn.kʼʷɨ.ra ‘egg’ kʷən.cɨ.wə ‘kind of small pot’ no.ʃa.ʃə ‘ram (male sheep)’ ɨm.me.tə.nə ‘mother of several young (goat/sheep)’ an.kʼə.fʷə ‘kind of spoon made of horn’ sɨ.βi.sa ‘piece of split bamboo to scrape stem of əssət’ yɨ.co.to ‘they farm, they work’ bə.su.βi ‘in the early morning’

(69) Word-final syllable CV# zɨ ‘this’ xɨ ‘that’ gʷə.la ‘pen inside the house for horses or mules’ əg.re ‘grassy place for cattle behind house (betw. gonə & river)’ ze.pə ‘central part of an old əssət’ tʼa.fi ‘kind of cereal (and the flour thereo)’ a.wə.do ‘white (used in combination with cereal or grain)’³⁵ cʼu.cʼu ‘chick’

(70) Word-initial syllable #CVC gʸɨm.mə ‘headrest of wood’ wɨs.sa ‘bread made from əssət’ kɨn.cɨf ³⁶ ‘beard’ gam.ba ‘kind of jar’ gen.zo ‘ax’ sən.kʼal.la³⁷ ‘small house (next to main house gʷəyə)’ jip.pə ‘mat (on floor inside house)’ gon.zɨ.yə ‘three earthen vessels on which cooking pot rests above fire’ tum.ma ‘garlic’

(71) Word-medial syllable CVC ɨn.kʼʷɨr.kʼʷi.nə ‘excrement of sheep or goat’ bə.tɨt.nət ‘width’ kʼʷɨr.cʼɨm.cʼɨm.yə‘ankle’ gɨ.ran.gɨr ‘young of a domestic animal’

of goat or sheep under the saddle or used to sit on’. ³⁵ According to Leslau (1979c: 112) the word is awədawə and means ‘kind of cereal’. He mentiones the form awədo for Endegeñ only. ³⁶ Leslau (1979c: 345) has, among others, kuncɨf for Chaha and kɨncɨf for Ezha. ³⁷ Leslau (1979c: 557) only has səkʼal(l)a without nasal.

31 kʷər.βeʃ.ʃa ‘young male sheep’ ɨn.dəx.re.cə ‘kidney’ a.fin.jə ‘(bird’s-eye) chili’ wə.zob.bɨd ‘have a mouthful of cʼat ()’ tʼɨ.rum.ba ‘trumpet’

(72) Word-final syllable CVC# ga.dɨr ‘cattle pen inside the house’ sɨ.kʼʷɨr ‘ceiling’ a.tʼɨm ‘bone’ də.rar ‘back part inside the house (~working place)’ ɨn.net ‘kind of bamboo’ wə.ʃər ‘small pot’ kʼʷɨ.rit ‘tip of house’ man.ʒor ‘aircraft’ nəβ.sə.tʼur ‘pregnant (humans)’

(73) Word-final syllable CVCC# an.ta.kɨrt³⁸ ‘eucalyptus’ wɨ.ra.wɨrx ‘what about the throat’ ʃɨn.kʷɨrt ‘onion’ a.rɨst ‘female (animal)’ yɨ.nə.kʼʷamt ‘Yinekwamt (subgroup of Gumer)’ an.ʃemxʷ ‘I did not hide’ yɨ.fəzz ‘it is better’ sə.jə.jimʃ ‘one drove him away, you know’ an.cʼorxʸ ‘you (s) did not carry’ tudxʸ ‘she tells you (s)’

2.3.2 Epenthesis 2.3.2.1 The epenthetic vowel ɨ The vowel ɨ is epenthetic with only few exceptions. It is used to split clusters of three or more consonants, but also CC word-initially and some CC word-finally. Apart from the two demonstratives zɨ ‘this’ and xɨ ‘that’, which end in ɨ, the exceptions concern lexemes beginning with this vowel. In particular, ɨ can be a trace of a lost consonant still present in related languages (for example ɨxa ‘water’, cf. Amharic wɨha; ɨmar ‘donkey’, cf. Arabic ħimaːr), or, following Banksira (2000: 156f.), when words begin with a geminate consonant (ɨyya ‘I’, ɨkka ‘like this’, ɨm- mat ‘only, one’). Further, there is a number of words with an initial ɨ followed by n (or an assimilated m) + plosive (ɨngʷəd ‘other’, ɨmbakʼʸə ‘saliva’, ɨmfas ‘air’). In

³⁸ Leslau (1979c: 104) has atankɨrt for ‘eucalyptus’ in Chaha.

32 all these cases the initial vowel ɨ is stable, i.e. like the other vowels it does not disappear when preceded, for example, by the prefix bə- ‘in (); with ()’ as in b-ɨngʷəd ‘in another; with another’; instead, the ə of the preposition is dropped. This is in contrast to verbs with the verbal prefix ɨn- (for example ɨnkʼɨrəkkʼər ‘move’, ↗ 3.5.3.4), the ɨ of ɨn- being dropped when preceded by another vowel (for example bə-nkʼɨrəkkʼərə ‘if it moved’). Finally, speakers occasionally realize an initial yɨ as ɨ resulting in variations like yɨna ~ɨna ‘we’ or yɨft ~ɨft ‘face, front’.

2.3.2.2 Epenthesis rules Rose (2000) discusses epenthesis in Chaha. The following is a summary of the general rules, which are the same in Gumer. Note that in addition to the expected forms as described below other outputs are possible. These rather marginal forms are indicated in brackets with a superscript inverted question mark (¿). Gumer allows for CV and CVC syllables (and a minimal syllable V word- initially, unless one assumes a glottal stop on its onset). Complex syllable onsets and codas are not possible, except some word-final CC# depending on the sonority of the involved consonants. Sonority also plays a role in determining the position of the epenthetic vowel ɨ that splits the prohibited *CCC clusters into CɨCC or CCɨC. Word-initially, all *#CC require epenthesis resulting in #CɨC. However, if the second C is r, the inserted ɨ is usually very short and hardly perceivable, which can be represented as #Cɨr.³⁹

(74) *#CC → #CɨC sɨma ‘listenǃ’ *#Cr → #Cɨr dɨrəkkərəm (~ dɨrəkkərəm) ‘he threw away’

Figure 5 shows the sonority hierarchy of consonants (labialized and palatalized consonants are not represented as they behave the same as their primary counterparts), which is relevant for the positioning of the epenthetic vowel in word-final CC# and CCC. Beginning with the glides, the consonants to the left have a higher sonority, while the stops to the right end rank lowest in the sonority hierarchy. There is no hierarchy among consonants of the same type. Fricatives and stops are obstruents, the remaining consonants being sonorants.

³⁹ Note that here the ɨ of #Cɨr is always written, for example zɨrəkkʸəm ‘he spoke’ (rather than zrəkkʸəm). Furthermore, also consider Banksira’s (2000: 25) statement that there is an epenthetic vowel in initial clusters, but in fast speech it is hardly audible, especially before sonorants, and that therefore he does not accept Cr clusters. This, however, is not completely in line with Banksira’s (2000: 181) footnote, where he mentiones that in fast speech a word-initial CC is not broken up by an epenthetic vowel if the second consonant is a sonorant, as for example in [grətəməm] ‘he cut in two’; this includes also β, for example [tʼβanərəm] ‘he folded’.

33 (w y) > r > β > m n > f s z x > t tʼ k kʼ d g (glides) > liquids > other > nasals > fricatives > stops son. sonorants > obstruents

Figure 5: Sonority hierarchy (based on Rose 2000: 405 / Chaha)

Word-final clusters CC# (75) are not split when the sonority of the first consonant is higher than that of the final one. In particular, (a) any sonorant followed by an obstruent and (b) obstruent clusters consisting of a fricative followed by a stop are not split. (The examples in (75) - (77) are all 2sm Imperatives.)

(75) falling S → CC# sonorant-obstruent zɨmd (¿zɨmɨd) ‘pullǃ’ fricative-stop kɨft (¿kɨfɨt) ‘openǃ’

In turn, when the sonority is rising or on the same level, word-final clusters are generally split (76). However, there is variation for the sequence r-sonorant as well as for obstruent-obstruent combinations, where speakers may also allow CC# (77).

(76) rising S → CɨC# obstruent-sonorant gətɨm (*gətm) ‘lend!’ equal S → CɨC# sonorant-sonorant gənɨm (*gənm) ‘give back loaned cow!’ (77) rising S → CɨC#~ CC# stop-frictative kɨtɨf ~ kɨtf ‘hashǃ’ equal S → CɨC# ~ CC# fricative-fricative məsɨx ~ məsx ‘chewǃ’ stop-stop əgɨd ~ əgd ‘tie!’ r-sonorant kʼɨrɨm ~ kʼɨrm ‘insult!’

Medial CCC clusters, i.e. sequences of three consonants between two vowels, have to be split into CCɨC or CɨCC. Here, the following syllable contact constraint plays a role (Rose 2000: 401): “The first segment of the onset of a syllable must be lower in sonority than the last segment in the immediately preceding syllable.” This constraint entails that in a CCC cluster with a rising-falling or a falling- rising sequence the epenthetic vowel has to split the consonants that feature a rise in sonority (forming a CɨC syllable) and thus allowing for the third consonant, which is part of the falling sequence, to become the onset of a following syllable, i.e. CɨC.C (78), or the coda of a preceding syllable, i.e. C.CɨC (79). The examples below all represent the Jussive stem CCC and the affixes for 3pm, i.e. yə-CCC-o. Note that they do not cover all possible combinations.

34 (78) rising-falling S → CɨCC obstruent-sonorant-obstr. yə.dɨm.do (¿yəd.mɨ.do) ‘let them gather’ stop-fricative-stop yə.kɨf.to (¿yək.fɨ.to) ‘let them open’ (79) falling-rising S → CCɨC sonorant-stop-fricative yən.kɨ.so (*yə.rɨk.so)⁴⁰ ‘let them bite’ fricative-stop-sonorant yəs.dɨ.βo (¿yə.sɨd.βo) ‘let them curse’

Medial CCC clusters with a consistent falling-falling sequence (80) do not violate the syllable contact constraint, thus the necessary epenthesis is equally possible in both positions. Here, the syllables are distributed according to an alignment rule saying that the maximal heavy syllables (CVC) pile up to the left of a word (cf. Rose 2000: 408ff.). Consequently, in a word of the shape yə-CCC-o the epenthetic vowel occurs between the second and third consonant resulting in yəC.Cɨ.Co (rather than yə.CɨC.Co).

(80) falling-falling S → CCɨC liquid-nasal-obstruent yən.fɨ.go (¿yə.rɨf.go) ‘let them be greedy’

However, medial CCC clusters with a consistent rising-rising sequence (81) violate the syllable contact constraint no matter where the epenthetic vowel is inserted. The theoretically possible solution with two epenthetic vowels (CɨCɨC) does not exist. Rather, also here the same alignment rule as above applies with the epenthetic ɨ between the last two consonants.

(81) rising-rising S → CCɨC obstruent-other son.-liquidyəs.βɨ.ro (¿yə.sɨβ.ro) ‘let them break’

Note that – as with the final clusters above –‘rising sonority’ also implies ‘equal sonority’. The examples in (82) represent an equal-rising and a falling-equal sequence, showing that they behave like (81) and (79) respectively.

(82) equal-rising S stop-stop-sonorant yəg.dɨ.ro (¿yə.gɨd.ro) ‘let them put to bed’ falling-equal S sonorant-stop-stop yən.kʼɨ.tʼo (*yə.rɨkʼ.tʼo) ‘let them kick’

There are also medial CCCC clusters, in particular with verbs that have a CCC Jussive stem followed by a suffix beginning with a consonant, for example 1p nɨ- CCC-nə. Generally, verbs with a rising-falling sequence (i.e. the ones with a CɨCC output of medial CCC clusters) show a CɨC.CɨC shape (83a), while all others verbs (with a CCɨC output of medial CCC clusters) result in the form C.CɨC.C (83b).

(83) a. nɨ.tʼɨβ.tʼɨ.nə (¿nɨtʼ.βɨtʼ.nə) ‘let’s grasp’ b. nɨs.dɨβ.nə (¿nɨ.sɨd.βɨ.nə) ‘let’s curse’

⁴⁰ r instead of n because it is not in penultimate coda position (↗ 2.1.3.3).

35 Again, it is important to note that although these rules predict the position of the epenthetic vowel in most instances, there is also idiolectal variation and other factors that can lead to different outputs⁴¹ (for more details see Rose 2000: 416ff.). The variation in final obstruent clusters is already exemplified above. Finally, according to Rose (2000: 415f.) and Banksira (2000: 26, 38), verbs (in Chaha) with two identical final consonants necessarily have an extra epenthesis between them to prevent the formation of a geminate. Such extra epenthesis is also found in Gumer, but contrary to Chaha it is not necessary: I have recorded many instances with two identical consonants not split by ɨ.

(84) kʼɨfɨf-o ~ kʼɨff-o ‘cut edgesǃ ( 2pm)’ yɨ-kʼɨfɨf-ʃə ~ yɨ-kʼɨff-ɨʃə ‘he will cut edges’ yə-sdɨd ~yə-sɨdd ‘may he chase away’ yɨ-rədɨd ~yɨ-rədd ‘it burns’ yɨ-gədɨd-o ~yɨ-gədd-o ‘they pierce’ yɨ-fəzɨz ~yɨ-fəzz ‘it is better’

2.3.2.3 Epenthetic vowel ɨ between words Clusters of three consonants across lexeme boundaries have to be split by means of the epenthetic vowel ɨ as well. This situation comes about when a word ending in CC is followed by a word beginning with C. This kind of epenthesis only takes place in continuous speech, i.e. there cannot be an epenthetic ɨ when the speaker interrupts the flow of the speech between the two words, be it a short hesitation or a longer pause. Nevertheless, also in uninterrupted speech the ɨ tends to be rather short up to the degree that often it is virtually not present. Most prominently ɨ is heard after words that end in doubled consonants (85), whereas after other clusters it is usually weaker (86), especially when r is involved (87). Note that in the following examples epenthetic ɨ is only indicated for illustrative purposes, but it is not written elsewhere.

(85) bɨrr ʃəm → [bɨrrɨ ʃəm] ‘he wanted birr/money’ ɨnkʼʷɨss barəm → [ɨnkʼʷɨssɨ barəm] ‘he shut up’ (86) yankʼ səβu → [yankʼɨ səβu] ‘he is the person of justice’ (87) yətɨmwərkʼ cʼɨnɨnya → [yətɨmwərkʼɨ cʼɨnɨnya] ‘Yetimwerk is in childbed’ The opposite, namely the lack of an epenthetic vowel due to a following word beginning in a vowel, can also happen. As is shown in (88), /tə-z-m/ on its own can only be realized with an epenthetic vowel between the two final consonants. In continuous speech, however, when /tə-z-m/ is followed by a word beginning with a vowel like /ema/, the whole phrase is resyllabified. With the m becoming the onset of a syllable, epenthetic ɨ is not necessary anymore.

⁴¹ For example yə.sɨβ.ro for yəs.βɨ.ro, yə.rɨm.do for yər.mɨ.do, nɨ.sɨd.βɨ.nə for nɨs.dɨβ.nə.

36 (88) /tə-z-m/ → [tə.zɨm] *[təzm] ABL-DEM-ALSO ‘(also) from this’ /tə-z-m ema/ → [təz.me.ma] ??[təzɨmema] ABL-DEM-ALSO way ‘(also) from this way’

2.3.3 Sequences of two vowels across morpheme boundaries When two vowels meet across morpheme boundaries there are basically five different possibilities how such a sequence is treated:⁴² - one vowel is deleted - the two vowels fuse to a different vowel - one vowel becomes a glide (i→y and u→w) - a glide (y or w) is inserted - nothing happens, i.e. the two vowels form a hiatus The central vowel ə is deleted when followed by e, o, ə, and a (89).⁴³ The same is true for the vowel a (90), but with a following e there is also the option of a hiatus or the insertion of a glide (d,e).⁴⁴

(89) a. agədə-e-m → agədem ‘he tied me’ b. tɨkə-o → tɨko ‘Oh childǃ’ c. tɨkə-əxʸta → tɨkəxʸta ‘her child’ d. tɨkə-axʸ → tɨkaxʸ ‘your (s) child’ (90) a. gʷeta-o → gʷeto ‘Oh Godǃ’ b. naβa-əxʸta → naβaxʸta ‘her waist’ c. naβa-axʸ → naβaxʸ ‘your waist’

⁴² Banksira’s (2000: 31f.) short description of the treatment of hiatus in Chaha served as a basis for this section (hence the mostly identical examples). Gumer and Chaha work differently in a few details but in some cases (in particular the ones involving glides) I am somewhat reluctant to believe that these are real differences in pronounciation. Probably this is only about differences in perception and/or notation. However the data may be incomplete and more variants exist (both among Gumer and Chaha speakers). ⁴³ In sequences of two identical vowels (i.e. ə+ə and a+a) it is actually futile to decide which element is deleted. In segmented and glossed examples I follow the practice of always leaving out the vowel in affixes: (i) tɨkə+əxʸta → tɨkə-xʸta vs. yə+əram → y -əram child-3sf -cow ‘her child’ ‘of cow’ (ii) abba+axə → abba-xə father-2sm ‘your father’

⁴⁴ In Chaha a+e results in ɛ (Banksira 2000: 31). At times examples like gəppem in (90d) could also be heard as gəppɛm, but generally in Gumer the vowel appears to be more closed.

37 d. gəppa-e-m → gəppem~ gəppaem ‘I understand’ ~gəppayem e. yɨ-gəβ-əma-e → yɨgəβəmae ‘in order for them to enter’⁴⁵ ~yɨgəβəmaye

The combination of ə+i fuses to e. This occurs particularly in the Impersonal of verbs with final radical A, for example with nəssa ‘lift’ and səmma ‘hear’ (91).

(91) a. nəʃʃə-i-m → nəʃʃem ‘one lifted him’ b. yɨ-səmʷə-i → yɨsəmʷe ‘one hears him’ Complete fusion to a different vowel occurs occasionally (in Chaha regularly) also with ə+u, a+u, and a+i, but normally the latter ones become the respective glides w and y (92).

(92) a. tɨkə-u → tɨkəw (~tɨko) ‘it is a child’ b. bora-u → boraw (~borɔ~boro) ‘it is an ox’ c. yɨ-fət-əma-i → yɨfətəmay ‘they () release him’ (~yɨfətəmɛ~yɨfətəme)

Final u and i change to the respective glides also after the vowels e and o.

(93) a. gʷəfəre-u → gʷəfərew ‘it is an Afro’ b. bʷɨtɨto-u → bʷɨtɨtow ‘it is worn out cloth’ c. yɨ-tʼəβtʼ-o-i → yɨtʼəβtʼoy⁴⁶ ‘they (m) grasp him’

The high vowels u and i, when followed by a non-high vowel, become the respective glides w and y, or alternatively the corresponding glide is inserted between the two vowels.⁴⁷

(94) a. yə-cəkkʷər-i-e→ yəcəkkʷərɨye ‘because one cooked it’ ~yəcəkkʷəriye b. wami-əta → wamyəta ‘his doing’ ~wamiyəta c. wami-axʸ → wamyaxʸ ‘your (s) doing’ ~wamiyaxʸ (95) a. tʼu-əta → tʼɨwəta~ tʼuwəta ‘his breast’ b. tʼu-axʸ → tʼɨwaxʸ ~ tʼuwaxʸ ‘your (s) breast’

⁴⁵ In the case of the purposive marker -e the more usual output after vowels is -y (for example yɨgəβəmay, cf. 4.7.3.6). They are treated here as allomorphs rather than a regular sound change since according to my observations it does not apply for 1s -e. ⁴⁶ Note, though, that 3pm -o + 3sm -i is often not realized as -oy. More commonly its form is either -əwi (i.e. with the o changing to əw), or oyi with the insertion of a glide (cf. example (401)). ⁴⁷ I do not have any examples for i+o. Banksira’s (2000: 31) yə-cəkkʷər-i-o → yəcəkkʷəryo ‘the cooked ones’ is not a suitable example since in my oppinion yo is not the result of two morphemes combined but simply the ‘heavy’ 3pm suffix -yo (↗ 3.12).

38 In case the second vowel is also high, I have usually noted variants where the second vowel changes to a glide (similar to the examples in (92) and (93)), but at least in some cases other outputs where the first vowel becomes a glide or with insertion of a glide are also attested.

(96) a. waβi-u → waβiw ‘he is generous’ b. bə-tʼu-u → bətʼuw ‘it is with breast’ c. y-acənə-xu-i → yacənəxuy ‘the one you brought’ ~yacənəxuyi d. sɨβi-i → sɨβyi~sɨβiyi ‘break (s) itǃ’

The mid-high vowels e and o followed by the central vowels ə (97a), (98a) or a (97b), (98b) form hiatus.⁴⁸ However, it seems that an o, similar to u+a → wa in (95b), can also become a glide w, but presumably only when preceded by a labializable consonant like k in (98c).

(97) a. gʷəfəre-əta → gʷəfəreəta ‘his Afro’ b. sɨne-axu → sɨneaxu ‘your (pm) wheat’ (98) a. asso-əna → assoəna ‘my salt’ b. bʷɨtɨto-axə → bʷɨtɨtoaxə ‘your (sm) torn clothes’ c. bʷɨllɨko-axʸ → bʷɨllɨkoaxʸ ‘your (s) toga’ ~bʷɨllɨkwaxʸ⁴⁹

Occasionally hiatus also seems to occur with other preceding vowels as with u +a as in (99), but it is probably better to conclude that in such cases the glide is extremely weak.

(99) a. cʼucʼu-axə → cʼucʼuaxə ‘your (sm) chick’ ~cʼucʼuwaxə Another borderline case where it is almost impossible to decide whether there is a hiatus or a glide represent the combinations o+e and a+e when a is not deleted. While I have usually perceived a hiatus, the insertion of a glide, which is very weakly pronounced though, is not wrong.

(100) a. tɨ-kʼəttʼɨn-no-e → tɨkʼəttʼɨnnoe ‘in order for her to kill them (m)’ b. yɨ-srə-βo-e → yɨsrəβoe ‘in order for them (m) to buy’ ~ yɨsrəβoye c. yɨ-gəβəma-e → yɨgəβəmae ‘in order for them () to enter’ ~ yɨgəβəmaye

⁴⁸ The portmanteau morpheme e- (< a-y- (-3)) before a base beginning with a occasionally seems to contain a weak glide, for example e-aʒ ~ ey-aʒ ‘he does not see’. When e+a is additionally preceded by a prefix the pronounciation can also be əy, for example təyaʒ ‘before he sees’. ⁴⁹ Alternatively, one might even analyze this output as labialized kʷ, i.e. bʷɨllɨkʷaxʸ.

39 Notice that all examples in (100) contain the suffixed purposive marker -e, which appears more commonly in its allomorphic form -y after any vowel (↗ 4.7.3.6), i.e. yɨsrəβoy rather than yɨsrəβoe ~ yɨsrəβoye. Finally, also the sequence of two e is realized with insertion of a hardly perceivable y.

(101) a. yɨ-məkʸr-e-e → yɨməkʸreye ‘so that it burns me’

2.3.4 Short note on stress According to Banksira (2000: 24), stress (in Chaha) falls on the penultimate syllable, no matter whether the syllable is open (102) or closed (103).

(102) acə́nəm ‘he brought’ acənə́xum ‘you (pm) brought’ (103) yə́ngɨd ‘let him touch’ yətmə́rkək ‘let him kneel down’

I can neither confirm nor deny this statement. In my impression the stress in words with an open penultimate syllable falls on the third but last syllable (for example ácənəm). But after all this is only an impression that cannot be proved here. Probably the difficulty in determining the stress bearing syllable lies in the question what exactly counts as stress: pitch or intensity. It could be the case that in Gurage languages these two features do not fall on the same syllable (Ronny Meyer, p.c.). As long as there is no further research on this topic, one has to go with Banksira’s native speaker intuition or knowledge.⁵⁰

⁵⁰ Some remarks on stress in Gurage languages can be found in Hetzron (1977: 42f.). Among several complications, his basic stress rule (based on Ennemor, i.e. Inor) differs from Banksira’s rule for Chaha. Thus, all in all, the situation is best summarized as not clear.

40 3 Verb morphology

3.1 Semitic root-and-pattern morphology Gumer features the typical root-and-pattern morphology common to all Semitic languages, which is well-known to linguists especially from Arabic. The different descriptions and studies of Semitic languages vary in their theoretical and practical treatment of this kind of morphology and the terminology used is not uniform, but generally speaking the differences are not significant. The following sections are not meant to discuss a new detailed analysis of Semitic morphology as a whole. Rather, they outline the Gumer system in a descriptive way comparable to previous studies of Semitic languages. Nevertheless, there will be a few alterations in the manner of representation that proved to be useful for Gumer. The basic principle of the root-and-pattern morphology is as follows. Firstly, there is a root, typically consisting of three consonants called radicals, which bears the lexical semantics; secondly, there are different regular patterns, also called templates, which determine the quality and position of vowels with respect to the radicals. Radicals on their own do not occur as actual words, but have to be integrated into a template to form a base. In the case of verbs, for example, these bases are then further affixed with subject markers resulting in complete minimal words or verb forms. For example, the root √kft consists of the three radicals k, f, and t, conveying the basic meaning ‘open’. Their position with respect to each other is fixed: k is the first radical (1), f the second one (2), and t the third one (3). As illustrated in Figure 6, to form the Perfective base the radicals are integrated into the template 1ə2ə3 yielding kəfət, whereas the pattern 1ə23 forms the Imperfective base kəft. The root √dmd ‘gather’, for example, fits into the same templates, resulting in the Perfective and Imperfective bases dəməd and dəmd respectively. When furnished with the 3sm subject markers, a suffixed -ə for the Perfective and a prefixed yɨ- for the Imperfective, these bases build the conjugated verb forms kəfətə/dəmədə and yɨkəft/yɨdəmd.

base| kəfət{z } | kəft{z }

verb form kəfət-ə yɨ-kəft

Figure 6: root/template/base/verb of √kft ‘open’

In Gumer, the root-and-pattern morphology is productive especially in the verb system; the formation of nominals based on this principle is more restricted.

41 For instance, there are no participles or nouns of place and time that would be formed productively with specific templates (and affixes) of potentially every single root. Rather, these concepts are expressed descriptively as “someone who is doing” or “place where one does”. Furthermore, there are no ‘broken’ or ‘internal’ plurals typical of other Semitic languages, since Gumer nominals do not form plurals at all.⁵¹ Nevertheless, there are nominals identifiable with roots also occurring in verbs,for example bʷɨsɨr ‘ripe’ and bəssər⁵² ‘cook, be ripe’ belonging to √βsr, or mʷɨzɨr ‘number’ and mezzər ‘count’ belonging to √mzr (cf. examples (622)-(626)). However, in the majority of cases the matching of root and template is not as straightforward as with the root √kft mentioned above. Due to many sound changes in the course of the history of the language and synchronic morphophonological processes, the relationship between words formed from the same root can be rather opaque. In particular due to the loss of some consonants, it is often not obvious anymore which template a surface word form (underlyingly) belongs to, which can, if at all, only be explained diachronically. For some verbs it is hardly possible to unequivocally assign them a root, as for example cona ‘sit’ and nɨwəʃʃə ‘get used to’. The former developed in an exceptional way from the reflexive tə- wəna (compare with the causative a-wəna ‘put’), the reflexive prefix tə- of which became palatalized to c (cf. Leslau 1979c: 559). Synchronically, the new root is best represented as √tUrA, but this structure does not exist elsewhere. The verb nɨwəʃʃə ‘get used to’ could be attributed to a root √rβsU or √rUsI (cf. Banksira 2000: 223). However, this verb is a loan from a Cushitic language (Leslau 1979c: 528), which is the main reason that it does not readily fit into the usual patterns (and also raises the question if every verb has to be attributed to a root).

3.2 Ethiosemitic verb types In studies on Ethiosemitic languages, verbs are traditionally classified into different verb classes or verb types. The basic types are called A, B and C (cf., for example, Hetzron 1972: 10); Rose (2007: 405f.), referring to Petros (1993), identifies an additional class D in Chaha. This classification is based on the absence or presence of gemination of the penultimate radical and the vowel quality in the templates of the three basic TAM forms Perfective, Imperfective and Jussive. The characteristics of the general Ethiosemitic verb types are: A gemination only in Perfective B gemination in all three TAM forms and presence of a palatal element C gemination in Perfective and Imperfective and vowel a after first radical The additional verb type D, which is not normally recognized in other Ethiosemitic languages, is characterized as: D gemination in all three TAM forms and presence of a labial element

⁵¹ Except for a few suppletive plural forms (see table 82). ⁵² Note that the citation form used here is  base (without the subject suffixes listed in table 56).

42 This classification concerns triradical verbs, i.e. the prototypical type of verb consisting of three radicals. In addition, there are also numerous verbs with four consonants called quadriradicals⁵³ (which traditionally are not assigned a specific label). Due to the diachronic loss of laryngeal consonants on the one hand, and the loss (or special behavior) of the glides w and y on the other hand, there is also a considerably high number of verbs occurring with less consonants in their roots (cf., for example, Ullendorff 1955: 24, 33, 35ff.; Hudson 1985: 39; Meyer 2011: 1244). Thus, in modern Ethiosemitic languages, one finds biradicals (and in Gumer even monoradicals) originating from triradicals, as well as tri- and biradicals originating from quadriradicals.⁵⁴

3.3 Gumer verb types Gumer verbs fall into the same Ethiosemitic verb types briefly outlined above. Due to the already mentioned fact that many verbs have undergone numerous morphophonological changes in the course of the history of the language, a more fine-grained classification is necessary to capture the various additional subtypes that have arisen in Gumer. In other words, one model verb does not suffice to explain all verbs that belong to a certain (basic) verb type, and further exempli- fications are needed. The two main morphophonological reasons responsible for these ʽirregularitiesʼ in the verb types are: (a) the diachronic loss of the post-palatal (velar, pharyngeal and glottal) consonants x, ħ, ʕ, ʔ and h on the one hand, as well as the loss of the approximants w and y on the other hand, the result of which is the missing of consonantal radicals in so-called ‘weak’ verb roots (as opposed to ‘sound’ verb roots); and (b) the status of etymological geminates of penultimate radicals. Below, the section 3.3.1 ‘Mutations’ discusses the problem of the etymological geminates. Then the regular (‘sound’) verb types are presented, followed by an illustration of the common ‘weak’ verbs, other special cases and exceptions.

3.3.1 Mutations As mentioned above, Ethiosemitic verbs geminate their penultimate radical in some forms as, for example, in the Perfective of type A verbs (but not in the Im- perfective or Jussive). Concerning this feature, the Gurage varieties have been divided into two groups, namely ‘geminating’ and ‘non-geminating’ (cf., for example, Leslau 1979c: lxxiii). Ezha belongs to the geminating languages, which have retained the etymological geminates. By contrast, in the non-geminating varieties like Chaha, the originally geminate penultimate radicals generally have been degeminated and occur as singletons. However, in most environments they

⁵³ Verbs with more than four radicals exist, but they are rather rare. ⁵⁴ In Leslau’s (1995: 280) Amharic grammar these verb types are called ‘abbreviated quadriradicals’ and ‘abbreviated triradicals’ or biradicals.

43 still show a reflex of the former gemination, i.e. where possible they are devoiced (for example g→k) or occlusivized (for example x→k). These sound changes are called here mutations. Figure 7 shows the pairs of basic consonants and their mutated forms.

basic mutated a. voiced fricative → voiceless fricative z, ʒ → s, ʃ b. voiced stop → voiceless stop β/b, bʷ → p, pʷ d, j → t, c g, gʷ, gʸ → k, kʷ, kʸ c. fricative → stop x, xʷ, xʸ → k, kʷ, kʸ d. liquid → nasal r → n

Figure 7: Mutation pairs

Consonants subject to mutation include the voiced consonants with voiceless counterparts, the fricative x/xʷ/xʸ, and the liquid r. The voiced consonants are devoiced, whereas x/xʷ/xʸ and r are occlusivized to k/kʷ/kʸ and n respectively. The nasal m/mʷ (n does not occur as basic root consonant) as well as the voiceless consonants cannot be altered.⁵⁵ Voiceless consonants include the ones that have no voiced counterpart (f and the ejectives), but note that also some of the voiceless consonants that occur as mutated forms can be the basic radicals. Thus for instance the t of kətəf ‘hash’ is original (√kt), whereas the t of sətəβ ‘curse’ is mutated (√sdβ). Gumer represents a mixed type concerning (de-)gemination.⁵⁶ In principle, it behaves like Chaha – there is degemination and the mutation patterns are the same –, but unlike Chaha degemination has not taken place in some cases. This leads to a slightly more complex situation than in ʽpureʼ non-geminating Gurage varieties (where in general mutation replaced gemination). Thus with the presence of gemination in some verbs in Gumer, there are four possible repre- sentations of an original geminate (in contrast to two possibilities in the non- geminating Chaha), summarized in table 1 with example verbs in the Perfective.

+ – – nəkəβ √rxβ ‘find’ nəgəd √rgd ‘touch’ + xəttər √xdr ‘thatch’ gʸəbbər √gβr ‘pay taxes’

Table 1: Mutation/gemination types

⁵⁵ The approximants w and y and the lateral l as penultimate radicals occur almost exclusively in loans and then only geminated, for example kʼəyyəs ‘plan’, dəwwəl ‘call, phone’, ella ‘covet’. ⁵⁶ Rose (2006) reports practically the same situation in Endegeñ; see Rose (2006: 845) for a proposed historical scenario of this development.

44 The occurrence of these four possibilities is not completely arbitrary. There is a considerable number of exceptions, but at large the distribution is as follows: –/+ The combination –gemination/+mutation occurs frequently and represents the “normal” case. For instance, the verb nəkəβ ‘find’, which is based on the root √rxβ, has a non-geminated but mutated penultimate radical (x→k). –/– The combination –gemination/–mutation only concerns verbs with penultimate radicals that can be devoiced, i.e. the absence of mutation in these verbs means absence of devoicing. Devoicing does not take place if the final radical is an obstruent (stop or fricative) except t (cf. Banksira 2000: 41ff., Rose 2006: 845). Examples are nəgəd ‘touch’ (√rgd) and nəgəf ‘drop, fall’ (√rg) without devoicing, as against sɨrəpət ‘spend some time’ (√srβt) with final radical t and devoicing (β→p). Mu- tation of r, i.e. occlusivization to n, is not affected by this rule, for example tənəf ‘remain’ (√tr). +/+ Gemination of the penultimate radical (be it mutated or not) occurs if the last radical is of short duration. In particular this is r, which is by far the shortest consonant.⁵⁷ Secondly gemination is also found with most verbs if the last consonant is missing (so to speak when it has a duration of 0 ms). Verbs with final radical r or missing final radical are surprisingly frequent (in contrast to the –gemination/–mutation verbs, which are significantly less numerous). Examples are xət- tər ‘thatch’ (√xdr) with a geminated and devoiced penultimate radical (d→tt), or bənnər ‘fly’ (√βrr) with a geminated and occlusivized penultimate radical (r→nn). +/– The combination +gemination/–mutation is attested in verbs like gʸəbbər ‘pay taxes, pay tribute’ (√gβr) or xədda ‘betray’ (√xdA). Since verbs belonging to this group are not very frequent and behave differently than comparable verbs like, for example, jəppər ‘finish’ (√dβr) and bətta (√βdA) ‘take’ without apparent reason (cf. Banksira 2000: 81f.), they might also be classified as exceptions (cf. Banksira 2000: 59).

3.3.2 Problem of (de-)gemination in Imperfective and Jussive So far, the problem of (de-)gemination has only been discussed for the Perfective, where the penultimate radical is subject to mutation in all verb types and occurs as geminate in the contexts described above, i.e. when the root final consonant is r or missing. However, depending on the verb type, mutation may also be re-

⁵⁷ Cf. measurements in Rose (2006: 846), where r has an average duration of 24.7 ms, compared to for example f with a duration of 106.3 ms.

45 quired in the Imperfective and/or Jussive. Here, the occurrence of geminates is more complicated and needs further elaboration. In a nutshell, the situation is best captured as follows. Geminating verbs also show gemination where required by the Imperfective and/or Jussive template, but only when the consonant in question is followed by a vowel. This vowel can be there as required by the template, but it can also be (and in terms of tokens often is) the epenthetic ɨ according to the rules of epenthesis (↗ 2.3.2). If the consonant in question is word final or directly followed by another consonant, i.e. if it appears in syllable coda position, it remains a singleton. For verbs with a final radical r, this means in particular that the singular forms show gemination, whereas the second and third person plural forms do not. This contrast is exemplified in (104) with some third person verb forms of type B verb √dβr ‘finish’.

(104)   3m  yɨ-jəppɨr yɨ-jəpr-o 3f  tɨ-dəppɨr yə-dəpr-əma

The number/gender suffixes -o (pm) and -əma (p) both begin with a vowel. When they are added to the base, the word is syllabified differently. The final r becomes the onset of a syllable and the preceding penultimate and potential geminate moves into the position of a syllable coda, where it appears as singleton only. It is important to note that the conceivable (or even expected) outcome *yɨjəppɨro (geminate + epenthetic vowel) does not occur (105).

(105) yɨ.jəp.pɨr + -o → yɨ.jəp.ro (*yɨ.jəpp.ro,*yɨ.jəp.pɨ.ro) ‘they finish’

Another instance is illustrated in (106). When the 3sm copula -u, consisting of a vowel only, is suffixed to an Imperfective with final r, gemination disappears as in the case above.

(106) tɨ.cək.kɨr + -u → tɨ.cək.ru (*tɨ.cəkk.ru,*tɨ.cək.kɨ.ru) ‘she is cooking’

When a suffix is added that begins with a consonant, as for example 1p -nə (107) or the definite future marker -te (108), the syllabification of the word is not altered. The final radical r remains in coda position, preceded by the epenthetic ɨ and the geminated penult.

(107) nɨ.jəp.pɨr + -nə→ yɨ.jəp.pɨn.nə (*nɨ.jəp.rɨ.nə) ‘we finish’ (108) yɨ.jəp.pɨr + -te → yɨ.jəp.pɨr.te (*yɨ.jəp.rɨ.te) ‘he will finish’

Note that the linker -m itself consists only of a consonant. However, when suffixed to a final r, epenthesis is needed. The resulting -ɨm works like a suffix beginning with a vowel (109).

(109) yɨ.jəp.pɨr + -m → yɨ.jəp.rɨm (*yɨ.jəp.pɨrm,*yɨ.jəp.pɨ.rɨm) ‘he finishes and’

46 As for the geminating verbs that lack a consonantal final radical, they have a Perfective base ending either in a or ə. The former also have a final a in the Im- perfective and Jussive. Consequently the penultimate radical is always followed by a vowel and thus geminated, i.e. of course only if required by the template (110).

(110) yɨkʼ.rəp.pa (*yɨkʼ.rə.pa) ‘he breaks at once, he snaps of’

The verbs with a base final ə in the Perfective do not show this vowel in the other TAM forms. Therefore the penultimate radical occurs word finally, constitutes a syllable coda and is not geminated (111).

(111) yɨz.rəkʸ (*yɨz.rəkkʸ) ‘he speaks’

According to the syllable structure rule of r-final verbs above, these verbs should show gemination when they receive a suffix that begins with a vowel. However, this is not the case. The penultimate radical remains a singleton, as in the singular form, irrespective of the type of suffix one adds, such as the number/gender marker -o (112a) or the linker -m (112b).

(112) a. yɨz.rəkʸ + -o → yɨz.rə.ko⁵⁸ (*yɨz.rək.ko) ‘they speak’ b. yɨz.rəkʸ + -m → yɨz.rə.kʸɨm (*yɨz.rək.kʸɨm) ‘he speaks and’

Finally, consider the following (113) minimal pair of the 3pm of the ə-final verb zɨrəkkʸə ‘speak’ and the a-final verb zɨrəkka⁵⁹ ‘spread, stretch’, showing that the former does not geminate whereas the latter does.

(113) a. yɨzrəkʸ → yɨzrəko ‘he speaks / they speak’ b. yɨzrəkka → yɨzrəkko ‘he spreads / they spread’

To sum up, gemination in verb forms is (largely) predictable. On the one side it can only be found where the verb template (i.e. morphology) allows for it, but its actual occurrence is conditioned by the environment, which is phonological in nature as described above. (Rose 2006: 847) explains the retained gemination (i.e. in Endegeñ, where the situation is basically the same) as “a compensatory trade-off between penultimate gemination and short final consonants and between penultimate singleton and long final consonants”. In other words, the total duration of the medial and final consonant remains more or less the same in all verbs; or put differently, gemination maintains the length of verbs that would be “too short” due to their short or missing final consonant. Further, also the fact that in the Imperfective and Jussive gemination does not occur in certain syllable constellations shows that gemination is a phonological phenomenon rather than a (pure) morphological one.

⁵⁸ With depalatalization kʸ→k (↗ 3.16). ⁵⁹ zɨrəkka is synonymous to zɨrakka, zɨrakka probably being the more common form.

47 The insight that the main purpose of the (retained) gemination is to compen- sate for the short duration of words leads to the assumption that gemination in longer words is less strong. And in fact, the longer a verb is or becomes with additional affixes, the weaker is the gemination. Thus, for example, the pp of səppərəm ‘he broke ()’ appears longer than the one in təsəppərom ‘they broke ()’, and in yɨtkʼʸəppərəma ‘they () receive’ it is hardly perceivable and virtually not geminated anymore. Yet geminated consonants are distinguishable from singletons, a fact which becomes clear in particular when geminated Gumer verbs are con- trasted with their non-geminated counterparts in Chaha. Since gemination is largely predictable, one might take into consideration not to write it. However, this requires that the non-native speaker (or in this case rather reader) knows which type a verb belongs to – at least for the verbs with a penultimate radical that cannot mutate. Compare, for instance, the similar looking verbs səffər ‘measure’ and cəffər ‘take a mouthful’. Both verbs have a final radical r and therefore are geminating verbs, but in the Imperfective only the latter one as a type B verb requires gemination, whereas the former one as a type A verb does not. Thus we have the contrast yɨsəfɨr ‘he measures’ vs. yɨcəffɨr ‘he takes a mouthful’. An example with verbs that lack a final consonantal radical is fətta ‘untie’ and təfətta ‘become loose’ with the corresponding Imperfective forms yɨfəta ‘he unties’ and yɨtfətta ‘it becomes loose’. Moreover, in addition to some exceptions, there are many Amharic loanwords in use that are pronounced with full gemination as in Amharic, even if the verb shape as such would predict a non- geminating verb according to the Gumer rules, an example being wəssən ‘decide’. There are also some minimal pairs of an original and a borrowed verb that contrast only in their gemination (in the Perfective), for example kʼənəs ‘begin’ vs. kʼənnəs ‘decrease’. For these reasons, gemination is represented in the notation wherever it occurs. Nevertheless, it must be noted that gemination in Gumer generally is not very strong and harder to perceive than, for example, in the related language Amharic. The matter is complicated by the fact that the strength of gemination can vary from speaker to speaker. On the one hand, one should take into consideration that there is a considerable number of ‘mixed’ marriages with one par- ent speaking a different Gurage variety. Thus, for example, a (non-geminating) Chaha mother can have a weakening influence, whereas the speech of a (geminating) Ezha mother might reinforce gemination. On the other hand, it has also been shown (Völlmin 2009) that gemination in Gumer appears slightly less often in villages close to the neighboring Chaha speaking area.

48 3.4 Main verb types 3.4.1 Sound verbs The bases of the triradical verb types A, B, C, D and the quadriradical verb types E and F are summarized in table 2. The radicals are represented with the numbers according to their position. The penultimate radicals that are subject to mutation (see section 3.3.1) are underlined ( ): they occur in the Perfective of all types, in the Imperfective of all types except type A, and in the Jussive of types B, C and D. Note that the Jussive of type C occasionally also appears without mutation.

Type Root   

A1 √123 1ə2ə3 1ə23 123 A2 √123 1ə2ə3 1ə23 12ə3 B √123 1ə2ə3 + 1ə23 + 1ə23 – C √123 1a2ə3 1a23 1a23 / 1a23 D √123 1ʷə2ə3 1ʷə23 1ʷə23 E √1234 12ə3ə4 12ə34 1ə234 F √1234 12a3ə4 12a34 12a34

Table 2: Verb types

3.4.1.1 Triradicals Type A Verb type A (table 3) is characterized by the presence of mutation of the penultimate radical in the Perfective and the absence thereof in the Imperfective and Jussive. There are two subtypes distinguished by the vowels in the Jussive: A1 without vowel (except for epenthetic ɨ, ↗ 2.3.2), and A2 with the vowel ə between radical 2 and 3. Verbs of subtype A1 are usually transitive and those of A2 intransitive. There are exceptions to this rule, as for example nəməd/rəmd/rməd ‘love’, which is transitive but belongs to subtype A2. Finally, note that verbs of type A2 often show variation and can alternatively also occur without ə in the Jussive.

Type Root   

A1 √srkʼ sənəkʼ sərkʼ srkʼ ‘steal’ A2 √zrβ zənəβ zərβ zrəβ ‘rain’

Table 3: Example bases of verb type A

The following lists show a choice of verbs belonging to type A1 (114) and A2 (115), grouped according to their patterns of gemination/mutation (↗ 3.3.1).

49 (114) Verbs of type A1 –/+ dənəg √drg ‘hit’ fənəm √frm ‘slice’ kʼənəm √kʼrm ‘insult’ kʼənəs √kʼrs ‘begin’ sənəβ √srβ ‘spin’ sənəkʼ √srkʼ ‘steal’ tʼənəkʼ √tʼrkʼ ‘scoop out’

–/+ (penultimate radical without mutated counterpart) bətəx √βtx ‘uproot, pull out’ kəfət √kft ‘open’ fətʼəm √ftʼm ‘close, block up’ tʼəkʼəm √tʼkʼm ‘be useful’ –/– nədəf √rdf ‘sting’ nəgəd √rgd ‘touch’ nəgəf √rgf ‘drop ()’ səgəd √sgd ‘pray, worship’ tʼəbəs √tʼβs ‘roast, fry’ tʼəβətʼ √tʼβtʼ ‘seize, hold’ zəgəd √zgd ‘remember’

–/– (penultimate radical without mutated counterpart) dəməd √dmd ‘meet’ kətəf √ktf ‘hash, chop’ kʼəməs √kʼms ‘taste’ nəfəg √rfg ‘be greedy’ nəfəkʼ √rfkʼ ‘tear off; hiccup’ nəkəs √rks ‘bite’ nəkʼətʼ √rkʼtʼ ‘kick, step on’ nətʼəkʼ √rtʼkʼ ‘snatch’ təkəs √tks ‘burn ()’ tʼəməd √tʼmd ‘yoke’ zəməd √zmd ‘pull’ +/+ dəppər √dβr ‘add’ kʼəppər √kʼβr ‘bury, plant’ məkkər √mxr ‘advise’ xəttər √xdr ‘thatch’

50 +/+ (penultimate radical without mutated counterpart) gəffər √gfr ‘push’ kʼəmmər √kʼmr ‘delouse’ kʼəttʼər √kʼtʼr ‘kill’ nəkkʼər √rkʼr ‘uproot, dig out’ səffər √sfr ‘measure’ səkkʼər √skʼr ‘hang’ tʼəkkʼər √tʼkʼr ‘hide ()’

(115) Verbs of type A2 –/+ fənəx √frx ‘tolerate’ gənəf √grf ‘be long’ sənəf √srf ‘be afraid, fear’ tənəf √trf ‘remain’ tʼənəkʼ √tʼrkʼ ‘be dry’ xənəm √xrm ‘stay a year’ zənəβ √zrβ ‘rain’

–/+ (penultimate radical without mutated counterpart) bəkət √βkt ‘perish, die (i.e. animals, without being slaughtered)’ –/– nəgəs √rgs ‘reign’

–/– (penultimate radical without mutated counterpart) nəməd √rmd ‘love’ nəfəs √rfs ‘blow (wind)’ +/+ bəkkər √βxr ‘lack’ bəttər √βdr ‘be first’ kʼəppər √kʼβr ‘be incomplete’ məkkər √mgr ‘suppurate’ nəppər √nβr ‘live’ səkkər √sxr ‘be drunk’ zəkkər √zgr ‘jump’

+/+ (penultimate radical without mutated counterpart) bəssər √βsr ‘cook (), be ripe’ fəkkʼər √fkʼr ‘be fat’ fəttʼər √ftʼr ‘be finished (food)’ nəttʼər √rtʼr ‘melt ()’

51 Type B Verb type B (table 4) is characterized by the mutation of the penultimate radical in all three TAM forms and in particular by the presence of a palatal element. This feature is represented as a superscript  after the first radical, i.e. √123.⁶⁰ The palatalization occurs on the first radical if it is a non-labial obstruent (t, tʼ, d, s, z, k, kʼ, g, x → c, cʼ, j, ʃ, ʒ, kʸ, kʼʸ, gʸ, xʸ). With all other radicals in the first position, the second radical is palatalized instead, but only if it is velar (k, kʼ, g, x → kʸ, kʼʸ, gʸ, xʸ). If neither the first nor the second radical is palatalizable, the first vowel is fronted to e (Rose 2007: 405). Palatalization only manifests in the Perfective and Imperfective; in the Jussive, the consonants appear depalatalized and the fronted vowels centralized, i.e. e→ə (↗ 3.16).

Type Root    B √dβr jəppər jə(p)pr də(p)pr ‘finish’ √mgr məkkʸər mə(k)kʸr mə(k)kr ‘burn’ √mtʼs metʼəs metʼs mətʼs ‘detach, break of’

Table 4: Example bases of verb type B

The list in (116) presents a choice of verbs belonging to type B, grouped according to their patterns of gemination/mutation (↗ 3.3.1).

(116) Verbs of type B –/+ cʼənəf √tʼrf ‘cut the top’ gʸənəz √grz ‘cut in big slices’ gʸətəm √gtm ‘lend’ ʃəpət √sβt ‘choose’ jəkəm √dgm ‘bash’

–/+ (penultimate radical without mutated counterpart) ʃəkət √skt ‘prepare, fix’ mesəx √msx ‘ruminate’ metʼəs √mtʼs ‘detach, break of’ +/+ cəkkər √tgr ‘cook ()’ jəppər √dβr ‘finish’ məkkʸər √mgr ‘burn, set fire’ ʃəkkər √sgr ‘change’

⁶⁰ Banksira (2000: 56ff.) analyzes these verbs as quadriradicals with a vocoid (radical) I after the first radical. Meyer (2011: 1244), discussing Gurage languages in general, mentions that type B exhibits a thematic vowel e ~ i between the first two radicals.

52 ʒəppər √zβr ‘return ()’

+/+ (penultimate radical without mutated counterpart) cəffər √tfr ‘take a mouthful’ gʸəttəβ √gtβ ‘bar’ mettʼər √mtʼr ‘select, pick’ nəkʸəm √rgm ‘ride, mount (a horse)’ +/– gʸəbbər √gβr ‘pay taxes’ mezzər √mzr ‘count’

Type C The main characteristics of verb type C (table 5) is the vowel a after the first radical in all three TAM forms. This feature is represented as a superscript , i.e. √123. Mutation of the penultimate radical occurs in the Perfective and Imperfective. In Ethiosemitic terms, the Jussive forms usually do not occur with mutation, but in Gumer they show variation in such a way that they mostly appear with mutation (cf. also Rose 2007: 408).

Type Root    C √mrx manəx manx manx / marx ‘capture’

Table 5: Example base of verb type C

The following list shows some examples of type C verbs (117), which are much less common than type A and B.

(117) Verbs of type C –/+ manəx √mrx ‘capture’ zapət √zβt ‘get lost’ ʒanəg √ʒrg ‘go away’⁶¹ +/+ kʼattʼər √kʼtʼr ‘knot’

Type D The characteristics of verb type D (table 6) are labialization of the first radical and mutation in all three bases.⁶² The labializing element is represented by the super-

⁶¹ ʒanəg/ʒang/ʒarg has a jussive variant zang with depalatalization, a feature of type B verbs. ⁶² Rose (2007: 406) states that type D has the same template in the Jussive as type B, but at the same time also that type D (unlike type B) does not feature mutation in the Jussive. This pattern could not be confirmed for Gumer where all verbs of type D show mutation in all bases.

53 script  after the first radical, i.e. √123.⁶³ If the initial radical is not labializable the first vowel is realized as o as it is the case in the verb zobbəd ‘have mouthful of cʼat’.⁶⁴ Note that there are other verbs with labialized consonants that do not belong to type D. Their labialization is primary and an inseparable part of the host radical and classify with one of the other types, for example type A tʼəkkʼʷər ‘become black √tʼkʼʷr, type B jəkkʷər ‘wilt, droop’ √dgʷr, type C zakkʼʷər ‘talk nonsense’ √zkʼʷr (cf. also Rose 2007: 406).

Type Root    D √kʼmr kʼʷəmmər kʼʷə(m)mr kʼʷə(m)mr ‘be strong’ √zβd zobbəd zobbd zobbd ‘have mouthful of cʼat’

Table 6: Example base of verb type D

Verbs belonging to this group are very rare. The list in (118) is probably close to exhaustive.

(118) Verbs of type D bʷənəs √βrs ‘feel lonely’ gʷənnər √grr ‘cut (in a special way)’ kʼʷəmmər √kʼmr ‘be strong, grow up’ kʼʷənnər √kʼrr ‘thatch peak of roof, trim, pile up’ zobbəd √zβd ‘have a mouthful of cʼatʼ

3.4.1.2 Quadriradicals The quadriradicals (table 7) mutate the penultimate radical in the Perfective and the Imperfective, but not in the Jussive. There are two types: type E, which is much more common, and type F. The latter is characterized by the vowel a – represented by a superscript , i.e. √1234 – which appears after the second radical in all three bases. By contrast, type E features the (default) vowel ə after the second radical in the Perfective and Imperfective only, whereas in the Jussive it appears after the first radical.

Type Root    E √grdm grətəm grətm gərdm ‘break in two’ F √frtx fratəx fratx fratx ‘mess’

Table 7: Example bases of quadriradicals

⁶³ Banksira (2000: 84) suggests that these verbs are quadriradicals with a vocoid (radical) U after the first radical, analoguous to type B verbs (Banksira 2000: 56ff.). For Eastern Gurage, Meyer (2011: 1244) suggests a verb class with a thematic vowel o ~ u between the first two radicals. ⁶⁴ Banksira (2000: 52) has the form zəbʷəd with labialization of the second radical for Chaha.

54 The following two lists show example verbs of type E (119) and type F (120). Note that type F is not very frequent.

(119) Verbs of type E dɨrəkkər √drgr ‘throw away’ gɨrətəm √grdm ‘break in two’ kʷɨrəkʷəm √kʷrkʷm ‘give blow with knuckles’ kʼɨrətʼəf √kʼrtʼf ‘pinch (finger)’ kʼɨrətʼəm √kʼrtʼm ‘step on member’ mɨsəkkər √mskr ‘testify’ sɨrəpət √srβt ‘spend some days’ xɨrətəm √xrtm ‘cut of’ (120) Verbs of type F fɨratəx √frtx ‘mess’ ʃɨraʃʃər √ʃrʃr ‘level the ground’ tʼɨβannər √tʼβrr ‘fold’

3.4.2 Weak verbs Weak roots lack one (or more) of their consonantal radicals. These missing radicals correspond to either the glides w or y, or to the post-palatal (velar, pharyngeal and glottal) consonants x, ħ, ʕ, ʔ, and h that have been lost in the course of the history of the language. Banksira (2000: 34) assumes as a placeholder for a missing laryngeal a vocoid A, and the vocoids U and I for the missing w and y respectively. This representation is adopted here. A usually surfaces as the vowel a, whereas U leads to labialization or vowel rounding and I to palatalization or vowel fronting.⁶⁵ It is important to observe that there are verbs occurring with three consonantal radicals on the surface that are not triradicals but originate from quadriradicals, for example fɨrəkkʼa from √frkʼA ‘split’ or zɨrəkkʸə from √zrgI ‘speak, talk’. The differing templates, in particular the position of the mutating radical, show that

⁶⁵ Compare these exemplary common Semitic (CS) roots (cf. Leslau 1987) with the actual Gumer equivalents: CS → Gumer Glides: √fsw ‘fart’ √fsU ‘fart’ fʷəʃʃəm ‘he farted’ √bky ‘cry’ √βxI ‘cry’ bəkkʸəm ‘he cried’ Post-palatals: √nfx ‘blow’ √rfA ‘blow’ nəffam ‘he blew’ √ftħ ‘open’ √ftA ‘untie’ fəttam ‘he untied’ √blʕ ‘swallow’ √βrA ‘eat’ bənam ‘he ate’ √mlʔ ‘be full’ √mrA ‘be full’ mənam ‘it was full’ √whb ‘give’ √Aβ ‘give’ aβəm ‘he gave’ Note, however, that in some cases these ‘weak’ consonants can vary from language to language. Compare, for example, Gumer fəccʼə ‘grind’ based on the root √ftʼI containing a final I (< glide y) with its cognates fasʼħa (√fsʼħ) ‘cut, split’ in Geez and fəssʼɨħe (√fsʼħ) ‘break to pieces’ in Tigrinya (Leslau 1987: 169) with a final post-palatal.

55 these verbs are not underlying triradicals. Consider, for example, the mutated kkʸ of zɨrəkkʸə, which is the last consonant on the surface but the penultimate radical of the root √zrgI.

3.4.2.1 Weak triradicals Table 8 summarizes the common weak root patterns of underlying triradicals, beginning with verbs featuring a weak first radical, followed by verbs with a weak penultimate (i.e. middle) radical and verbs with a weak final radical. Note that most weak triradicals are attributable to type A.

Type/subtype   

A1-U23 wə2ə3 wə23 (w)23 A2-U23 wə2ə3 wə23 (w)2ə3 A1-A23 a2ə3 a23 (ə)23 A2-A23 a2ə3 a23 (ə)2ə3 A-1U3-w 1ʷə3 1ʷə3 1ʷ3 A-1U3-o 1o3 1o3 1o3 B-1U3 1o3 + 1o3 + 1o3 - A-1A3-ə 1a3 1ə3 1a3 A-1A3-Ø 1a3 1(ʸ)3 1a3 A-1I3-y 1ʸə3 1ʸə3 1ʸ2 A-1I3-e 1e3 1e3 1e3 / 1i3

A1-12U-w1 1ʷə2ə 1ʷə2 1ʷ2 A2-12U-w2 1ʷə2ə 1ə2ʷ 12ʷe C-12U 1ʷa2ə 1ʷa2 1ʷa2 A-12A 1ə2a 1ə2a 12a A1-12I-y 1ə2ʸə 1ə2ʸ 12ʸ A1-12I-e 1e2ə 1e2 1i2 A2-12I 1ə2ʸə 1ə2ʸ 12e B-12I 1ə2ə + 1ə2 + 1ə2 - C-12I 1a2ʸə 1a2ʸ 1a2ʸ

Table 8: Common patterns of weak triradical roots

3.4.2.1.1 Triradicals with weak first radical Type U23 Triradicals with U as the first radical (table 9) do not lack a consonant in the Perfective and Imperfective, but surface with w. In the Jussive, however, this w is often dropped after the subject prefixes if the verb is intransitive (i.e. type A2),

56 whereas transitive verbs (i.e. type A1) usually do not elide w (cf. Rose 2007: 408).⁶⁶ In the Imperative, i.e. when there is no preceding subject marker, w is present in all cases.

Type Root   

A1-U23 √Ukʼs wək’əs wək’s wk’s ‘blame’ A2-U23 √Utʼkʼ wət’ək wət’k’ (w)t’ək’ ‘fall’

Table 9: Example bases of triradical verbs with root U23

(121) Verbs of type A1-U23 wəkʼəs √Ukʼs ‘blame’ wəsəd √Usd ‘take, take away’ wəttər √Utr ‘draw tight’

(122) Verbs of type A2-U23 wətʼəkʼ √Utʼkʼ ‘fall’

Type A23 Triradicals that feature A as the first radical (table 10) start with the vowel a, followed by radicals 2 and 3. In the Jussive a changes to ə, which is elided after all subject prefixes but is still present in the Imperatives.

Type Root   

A1-A23 √Arβ anəβ arβ (ə)rβ ‘milk’ A2-A23 √Adr attər adr (ə)dər ‘spend the night’

Table 10: Example bases of triradical verbs with root A23

(123) Verbs of type A1-A23 adəg √Adg ‘throw down, make fall’ afəs √Afs ‘scoop with both hands’ agəd √Agd ‘tie, bind’ agəz √Agz ‘help’ anəβ √Arβ ‘milk’ atʼəβ √Atʼβ ‘wash’ attʼər √Atʼr ‘make a fence’

⁶⁶ wəsəd ‘take’, for example, is an exceptional A1 verb that elides the w in : wəsəd / wəsd /(w)sd.

57 (124) Verbs of type A2-A23 ammər √Amr ‘believe’ anəkʼ √Arkʼ ‘be finished, come to an end’ anəs √Ars ‘be less, be smaller’ attər √Adr ‘spend the night’ attʼər √Atʼr ‘be short’

Type I23 Unlike w, there are no verbs with an initial consonant y. The few verbs beginning with e might be attributed to verb type I23 (cf. Rose 2007: 409). However, they are treated here as exceptions; see section 3.5.3.10 for examples.

3.4.2.1.2 Triradicals with weak second radical Type 1U3 Triradicals with U in the second radical position (table 11) occur with only two consonants, 1 and 3. Since it is the penultimate radical that is missing, mutation does not apply. U labializes the first radical if it is labializable, i.e. velars and labials (A-1U3-w). Otherwise it is realized as vowel o with verbs belonging to type A (A- 1U3-o) and verbs with palatalization belonging to type B (B-1U3).

Type Root    A-1U3-w √fUx fʷəx fʷəx fʷx ‘wipe’ A-1U3-o √tʼUm tʼom tʼom tʼom ‘fast’ B-1U3 √tUt cot cot tot ‘work’

Table 11: Example bases of triradical verbs with root 1U3

(125) Verbs of type A-1U3-w fʷəx √fUx ‘wipe’ kʼʷəm √kʼUm ‘stand’ mʷət √mUt ‘die’

(126) Verbs of type A-1U3-o notʼ √rUtʼ ‘run’ tos √tUs ‘be confused’ tʼom √tʼUm ‘fast’

58 (127) Verbs of type B-1U3 cot √tUt ‘work’⁶⁷ cʼor √tʼUr ‘carry (loads)’ jos √dUs ‘bash, hit with fist’ ʒor √zUr ‘go round’

Type 1A3 Triradicals with a second radical A (table 12) occur only with two consonants, 1 and 3. Since A stands for the penultimate radical that is missing, mutation does not apply. Typically, A is realized as a in the Perfective and Jussive, and as ə in the Imperfective (A-1A3-ə) (128). A few verbs do not have a vowel in the Imperfective (A-1A3-Ø) (129), but instead their first radical is palatalized when palatalizable. This is the case with the two verbs xar ‘know’ and sar ‘be pleasing, be pretty’ whose Imperfective bases are xʸr and ʃr respectively (b). The verb cal ‘can’, a frequently-used loan from Amharic, also belongs to the latter group (c). The very frequent verb bar ‘say’ (d) is irregular: its Jussive features the vowel ə (↗ 3.17).

Type Root    A-1A3-ə √dAkʼ dakʼ dəkʼ dakʼ ‘laugh’ A-1A3-Ø √βAs βas βs βas ‘be bad’ √sAr sar ʃr sar ‘be pleasing’ √βAr βar βr βər ‘say’

Table 12: Example bases of triradical verbs with root 1A3

(128) Verbs of type A-1A3-ə cʼar √cʼAr ‘load; saddle’ dakʼ √dAkʼ ‘laugh’ fakʼ √fAkʼ ‘scrape’ gad √gAd ‘be hungry’ gar √gAr ‘be red-hot’ kas √kAs ‘divide’ kʼam √kʼAm ‘chew cʼat’ mʷakʼ √mʷAkʼ ‘be warm’ mar √mAr ‘have pity’ nax √rAx ‘send’ naz √rAz ‘be heavy’

⁶⁷ Note that this verb developped from a derived form of a root √Ut (Leslau 1979c: 175), i.e. tə-wat, the detransitivizer tə- of which became palatalized to c. This verb is to be identified with words like wat ‘digging, ploughing; ploughed field’ and cəwac ‘farmer’.

59 sam √sAm ‘kiss’ ʃaβ √ʃAβ ‘pull’ tam √tAm ‘be sweet, taste good’ tʼaf √tʼAf ‘write’ war √UAr ‘spend the day’ watʼ √UAtʼ ‘swallow’

(129) Verbs of type A-1A3-Ø a. bas √βAs ‘be bad’ nakʼ √rAkʼ ‘be bigger, surpass’ b. sar √sAr ‘be pretty, be pleasing’ xar √xAr ‘know’ c. cal √cAl ‘can’ (< ) d. bar √βAr ‘say’ (irregular, ↗ 3.17)

Type 1I3 Triradicals with I as second radical (table 13) occur with two consonants, 1 and 3. Since I is in the position of the penultimate radical, mutation does not apply. The element I causes palatalization of a preceding consonant (A-1I3-y) (130) or it is realized as vowel e (A-1I3-e) (131).⁶⁸

Type Root    A-1I3-y √kʼItʼ kʼʸətʼ kʼʸətʼ kʼʸtʼ ‘be tired’ A-1I3-e √tIg teg teg teg / tig ‘be done successfully’

Table 13: Example bases of triradical verbs with root 1I3

(130) Verbs of type A-1I3-y kʼʸətʼ √kʼItʼ ‘be tired’

(131) Verbs of type A-1I3-e ʃem √ʃIm ‘hide’ teg √tIg ‘be done successfully’ ter √tIr ‘be late (time)’

⁶⁸ I have no explanation for verbs of type A-1I3-e like teg ‘be done successfully’; the initial t is a palatalizable consonant, yet it is not palatalized.

60 3.4.2.1.3 Triradicals with weak third radical Type 12U Triradicals with a third radical U (table 14) occur only with two consonants, 1 and 2. The element U labializes the rightmost labializable radical (i.e. velars and labials) preceding it and causes palatalization of 2 if it is palatalizable (i.e. alveolars) (cf. Banksira 2000: 22). Note that the final ə in the Perfective is part of the base (i.e. suffixed subject markers follow the ə, ↗ 3.11.6.1). The verb tʼəffʷə ‘be satiated, full’ seems to be the only example belonging to the type A2. Banksira (2000: 202) mentions that (in Chaha) “[t]he stem-final /U/ of /t’ɨfəU/ exceptionally palatalizes /ə/ to [e] only in the Imperative”, i.e. the base is tʼfʷe with a final e. Thus, the reason for the unusual vowel e lies in the shape of the template of A2 (similar to səccʼə ‘drink’ below), where there is the vowel ə between the penultimate and last radical, i.e. underlyingly we have tʼfəU. While the labial f is labialized by U (and not palatalized), the interfering ə is subject to the palatalizing effect of U and becomes fronted to e. Note, however, that in the Perfective there is also a vowel ə, which is not palatalized (cf. Banksira 2000: 202). Nevertheless, in Gumer I have also recorded tʼfʷ without final e (both for Imperatives and the other Jussive forms).⁶⁹ Presumably the alternative base exists because occasionally the Jussive of verbs of type A2 can also be formed according to the template of type A1 which lacks the vowel ə (↗ 3.4.1.1).

Type Root   

A1-12U-w1 √fsU fʷəʃʃə fʷəʃ fʷʃ ‘fart’ A1-12U-w2 √rkʼU nəkkʼʷə rəkʼʷ nkʼʷ ‘roar’ A2-12U-w2 √tʼfU tʼəffʷə tʼəfʷ tʼfʷe~ tʼfʷ ‘be satiated, full’ C-12U √ftʼU fʷaccʼə fʷacʼ fʷacʼ ‘mow’

Table 14: Example bases of triradical verbs with root 12U

As examples (132)-(135) show, triradicals with root 12U are not very frequent.

(132) Verbs of type A1-12U-w1 kʼʷənə √kʼrU ‘roast’ fʷəʃʃə √fsU ‘fart’

(133) Verbs of type A1-12U-w2 nəkkʼʷə √rkʼU ‘roar’

⁶⁹ It is not clear whether in Chaha only the Imperative features this final e or all Jussive forms.

61 (134) Verbs of type A2-12U-w2 tʼəffʷə √tʼfU ‘be satiated, full’

(135) Verbs of type C-12U fʷaccʼə √ftʼU ‘mow’ mʷaʃʃə √msU ‘rub, massage’

Type 12A Triradicals with A on the third radical position (table 15) occur with the two consonants 1 and 2 and a final a, which belongs to the base. For this rather frequent type there are no formal subtypes A1 and A2 since the distinctive feature, the occurrence of the vowel ə in the Jussive of the latter, is neutralized by the final a.

Type Root    A-12A √gβA gəppa gəβa gβa ‘enter’

Table 15: Example base of triradical verb with root 12A

(136) Verbs of type A-12A anna⁷⁰ √ArA ‘defecate’ bəna √βrA ‘eat’ bətta √βdA ‘take’ fəkka √fkA ‘escape’ fəkkʼa √fkʼA ‘split’ fəna √frA ‘have intercourse’ fətta √ftA ‘untie’ gəffa √gfA ‘push’ gəppa √gβA ‘enter’ gəssa √gzA ‘own, govern’ gətta √gdA ‘pour’ kəna √krA ‘ascend’ kəppa √kβA ‘bend, fold’ kʼəppa √kʼβA ‘smear, oil’ kʼəttʼa √kʼtʼA ‘punish’ məna √mrA ‘be full’ nəppa √rβA ‘split’ nədda √rdA ‘help’ nəffa √rfA ‘blow’ nəkka √rgA ‘coagulate’ nəmma √rmA ‘grow, sprout’

62 nəssa √rsA ‘lift, raise’ nətta √rtA ‘separate (layers of əssət)’ səmma √smA ‘hear’ səna √srA ‘arrive, reach’ ʃətta √ʃtA ‘smell ()’ təffa √tfA ‘slap, spit’ təppa √tβA ‘be hard’ tʼəppa √tʼβA ‘skin’ tʼəmma √tʼmA ‘be thirsty’ tʼəna √tʼrA ‘call; hate; be expensive’ tʼəffa √tʼfA ‘be extinguished’ wəkka √UgA ‘stab’ wəttʼa √UtʼA ‘go up, go out’ xədda √xdA ‘betray’ xəna √xrA ‘put; forbid’ zəna √zrA ‘sow’

Type 12I Triradicals with a third radical I (table 16) occur only with two consonants, 1 and 2. I palatalizes the directly preceding penultimate radical if it is palatalizable or otherwise raises the first vowel (ə → e or ɨ → i). In the case of kʼʸəmmə ‘defeat’ the initial velar is palatalized. This is probably the only such example. In contrast, initial velars of verbs with penultimate radical r like xənə ‘dig a hole’ do not palatalize even though the mutated n in the Perfective is not palatalizable (the non-mutated r in the Imperfective and Jussive on the other hand is palatalized regularly, cf. table 73). Note that the final ə in the Perfective is part of the base (i.e. suffixed subject markers follow the ə). Verbs of type A2 that have a palatalized penultimate in the Perfective and Imperfective, like səccʼə ‘drink’, appear with a depalatalized consonant but a final e instead in the Jussive (stʼe). This can be explained by the template of A2, where there is a vowel ə between the penultimate and the final radical, i.e. underlyingly we have stʼəI with the interfering ə “ab- sorbing” the palatalization to become fronted to e. Nevertheless, note that in the Perfective there is also a vowel ə, which is not fronted.⁷¹

⁷⁰ In contrast to the singleton n in verbs with penultimate radical r and weak final radical, anna ‘defecate’ shows geminated nn, possibly explainable as compensation for the initial weak radical A (compare with the discussion of retained gemination on page 47). ⁷¹ Furthermore, due to the fact that in principle all verbs of type A2 can occur with the Jussiv template of A1 without vowel ə, the Jussive base of √stʼI is also attested as scʼ (instead of stʼe), i.e. the final I directly palatalizes tʼ. Thus, for example, the Imperative 2sm is not only sɨtʼe but also sɨcʼ.

63 Type Root   

A1-12I-y √sxI səkkʸə səxʸ sxʸ ‘flee’ √kʼmI kʼʸəmmə kʼʸəm kʼʸm ‘defeat’ A1-12I-e √sfI seffə sef sif ‘sew’ A2-12I √stʼI səccʼə səcʼ stʼe ‘drink’ B-12I √trI cənə cən tən ‘come’ C-12I √kʼsI kʼaʃʃə kʼaʃ kʼaʃ ‘throw away’

Table 16: Example bases of triradical verbs with root 12I

(137) Verbs of type A1-12I bəkkʸə √βxI ‘cry’ fəccʼə √ftʼI ‘grind; sharpen’ məʃʃə √msI ‘become evening’ nəccʼə √rtʼI ‘pluck’ səkkʸə √sxI ‘flee’ kʼʸəmmə √kʼmI ‘win, defeat’

(138) Verbs of type A1-12I-e seffə √sfI ‘sew’ cʼeffə √cʼfI ‘dip, soak’

(139) Verbs of type A2-12I səccʼə √stʼI ‘drink’ akkʼʸə √AkʼI ‘crunch grain, chew’

(140) Verbs of type B-12I cənə √trI ‘come’ cʼənə √tʼrI ‘give birth’

(141) Verbs of type C-12I gaʃʃə √gsI ‘make a raid’ kʼaʃʃə √kʼsI ‘throw away’ maccə √mtI ‘be angry’

3.4.2.2 Weak quadriradicals Table 17 summarizes the most common weak root patterns of underlying quadriradicals, i.e. those roots with a missing final radical (other types are rare). These

64 verbs behave like the weak triradicals with a missing final radical. In particular, this means that the radical that is subject to mutation appears as the last consonant on the surface. Note that quadriradicals of type F are rare.

Type    E-123U 12ə3ə 12ə3 1ə23 E-123A 12ə3a 12ə3a 1ə23a F-123A 12a3a 12a3a 12a3a E-123I 12ə3ə 12ə3 1ə23 F-123I 12a3ə 12a3 12a3

Table 17: Common patterns of weak quadriradical roots

Type 123U Quadriradicals with U as last radical (table 18) occur with three consonants: 1, 2 and 3. The element U labializes the rightmost labializable radical (velars and labials) preceding it and causes palatalization of radical 3 if it is palatalizable (i.e. alveolars). Note that the ə in the Perfective is part of the base. No examples for type F (with vowel a) could be found.

Type Root    E-123U √tʼrkʼU tʼrəkkʼʷə tʼrəkʼʷ tʼənkʼʷ ‘be dea’ √kʼrtʼU k’ʷrəccʼə kʼʷrəcʼ kʼʷəncʼ ‘take a handful’

Table 18: Example bases of quadriradical verbs with root 123U

(142) Verbs of type E-123U fɨkʼʷənə √fkʼrU ‘whistle’ kʼʷɨrəccʼə √kʼrtʼU ‘take a handful’ tʼɨrəkkʼʷə √tʼrkʼU ‘be dea’ zɨwənə √zβrU ‘envy’

Type 123A Quadriradicals with A as last radical (table 19) occur with three consonants: 1, 2 and 3. All forms have a final a, which belongs to the base.

65 Type Root    E-123A √kʼrβA kʼrəppa kʼrəppa kʼəmba ‘pinch of’ F-123A √zrgA zrakka zrakka zraga ‘stretch out’

Table 19: Example bases of quadriradical verbs with root 123A

(143) Verbs of type E-123A fɨrətta √frtA ‘scatter, sprinkle’ fɨrəkkʼa √frkʼA ‘hatch’ gɨrəppa √grβA ‘bend ()’ kʼɨrəffa √kʼrfA ‘hit with stick or fist, parry’ kʼɨrəppa √kʼrβA ‘break at once, snap of’

(144) Verbs of type F-123A zɨrakka √zrgA ‘stretch out’

Type 123I Quadriradicals with I as last radical occur with three consonants: 1, 2 and 3. The element I palatalizes the preceding consonant if it is palatalizable. Note that the ə in the Perfective is part of the base.

Type Root    E-123I √kʼmtʼI kʼməccʼə kʼməcʼ kʼəmcʼ ‘be ashamed’ F-123I √βrtʼI βraccʼə βracʼ βracʼ ‘scatter’

Table 20: Example bases of quadriradical verbs with root 123I

(145) Verbs of type F-123I dɨβəʃʃə √dβsI ‘pound root of əssət’ fɨrəkkʸə √frkI ‘break in two’ kʼɨməccʼə √kʼmtʼI ‘be ashamed’ zɨrəkkʸə √zrgI ‘speak’

(146) Verbs of type F-123I bɨraccʼə √βrtʼI ‘scatter’

66 Other quadriradical types There are virtually no quadriradicals with a weak radical other than the final one. I am aware of two instances, nɨwəʃʃə ‘get used to’, which is based on either the root √rβsU or √rUsI (cf. Banksira 2000: 223), and sɨmat ‘urinate’, which seems to have the root √smAt. Nevertheless, it is worth remarking that both are loans from Cushitic languages (Leslau 1979c: 528, 579).⁷² This fact explains their unusual structure and it is questionable if they should be identified with a root at all. Nevertheless, as illustrated in table 21, they can be fit into the templates of type E. There is gemination in the Perfective and Imperfective, but of course this is not visible in the case of sɨmat where the penultimate radical is not a consonant. Further, note that sɨmat is similar to verbs of type A-1A3-ə (see above), which change their penultimate vowel from a to ə in the Imperfective, but with an additional initial consonant s. The vowel o of the Jussive base of nɨwəʃʃə is the result of əw (the template being 1ə23 → *rəwʃ → roʃ ).

Type Root    E √rUsI/rβsU nwəʃʃə rwəʃ roʃ ‘get used to’ √smAt smat smət smat ‘urinate’

Table 21: Bases of nɨwəʃʃə ‘get used to’ and sɨmat ‘urinate’

3.5 Special verb classes, subtypes and exceptions 3.5.1 Monoradicals Table 22 summarizes the rare monoradicals, i.e. verbs that surface with only one (initial) consonantal radical (for aβ ‘give’ and od ‘tell’ see section 3.5.3.7). The verbs ʃə ‘want’ and cə ‘leave’ belong to type B with a palatal element in the Per- fective and Imperfective and depalatalization in the Jussive. The verb xʷə ‘spill’ features labialization and might be attributable to type A or D. All three verbs have a diphthong ay in the Jussive, but cə differs from the other two in that it has no vowel in the Imperfective.

Type    B ʃə ʃə say ‘want’ cə c tay ‘leave’ ― xʷə xʷə xʷay ‘spill’

Table 22: Monoradicals

⁷² sɨmat ‘urinate’ seems to be denominative from sɨmat ‘urine’, which itself comes from a Cushitic verb root ʃumaʔ- (or similar) furnished with a (Gurage) nominal suffix -(a)t (Leslau 1979c: 579f.).

67 3.5.2 Reduplicated verbs Usually, verb roots consist of three (or four) distinct radicals. In some verb roots, however, two radicals are identical, i.e. reduplicated. Usually the following three different types of reduplication in verb roots are distinguished (cf. Banksira 2000: 37, Rose 2007: 409): (a) final reduplication, i.e. triradicals and quadriradicals with an identical penultimate and final radical and the patterns 122 and 1233 respectively; (b) total reduplication, i.e. quadriradicals with the pattern 1212; (c) medial reduplication, i.e. verbs with the pattern 1223. Medial reduplication is commonly known in Ethiosemitic languages as the Frequentative. This form has to be distinguished from the other two reduplica- tions in that it is a regular (more or less productive) derivation that adds the notion of intensity, repetition or distribution to the meaning of the basic (triradical) verb. Frequentatives are dealt with separately in section 3.10. In contrast, final and total reduplication concern the special appearance of some verb forms and are described in the following paragraphs.

3.5.2.1 Final reduplication (122 / 1233) Triradical verbs with final reduplication have the shape 122. They do not convey a specific meaning. Some authors regard them as biconsonantal roots that double their second radical to conform to the common triradical pattern (cf. Rose 2007: 409), but here they are treated as a special case of triradicals.⁷³ They belong to all basic verb types and their penultimate radical (i.e. the first of the identical two radicals) is subject to mutation where required. table 23 shows the bases of example verbs of each type.

Type Root    A √βrr bənnər bərr brər ‘fly’ B √βtt betət bett bətt ‘be wide’ C √βzz bazəz bazz bazz ‘feel depressed’ D √kʼrr kʼʷənnər kʼʷə(n)nr kʼʷə(n)nr ‘trim, pile up’

Table 23: Example bases of triradicals with final reduplication

Like triradicals with final reduplication, also quadriradicals with identical final radicals are not derivations. However, according to Prunet & Petros (1996), they often convey a meaning of repetition, “local movements” (actions close to the body or small repetitive movements), or physical impairment (the same is true for quadriradicals with total reduplication, ↗ 3.5.2.2). This type of verb has

⁷³ Thus, the verbs bənnər ‘fly’ or fəzəz ‘be better’, for example, can be seen as having the root √βr or √βrr and √fz or √fzz, respectively.

68 the shape 1233 but behaves like ordinary quadriradicals.⁷⁴ Table 24 lists the bases of example verbs of the two quadriradical verb types E and F.

Type Root    E √kʼmtʼtʼ kʼmətʼətʼ kʼmətʼtʼ kʼəmtʼtʼ ‘wrinkle’ F √tʼβrr tʼβannər tʼβa(n)nr tʼβarr ‘wrap up, fold’

Table 24: Example bases of quadriradicals with final reduplication

As mentioned above, all verbs with final reduplication behave in principle like ordinary tri- and quadriradicals. Thus, it is only the penultimate radical that is mutated where required, but not the doubled final radical, for example bənnər ‘fly’ (√βrr) or jəpəβ ‘dam’ (√dββ)⁷⁵. Nevertheless, palatalization and labialization of the last radical affects usually⁷⁶ both consonants of the doubled pair. Palatalization occurs in the feminine singular in the Imperfective and Jussive (147) (↗ 3.11.4), labialization in any verb form with the light 3sm suffix -n (148) (↗ 3.12.1), and palatalization or palatalization in all Impersonal forms (149) (↗ 3.11.5).

(147) tɨgədd ‘you (sm) pierce’ → tɨgəjj ‘you (s) pierce’ afʷkɨk ‘squat (sm)!’ → afʷkʸɨkʸ ‘squat (s)!’ (148) kʼəfəfəm ‘he cut (fingernails)’ → kʼəfʷəfʷənɨm ‘he cut it’ (149) fəzəzəm ‘he got well’ → fʷəʒəʒim ‘one got well’ kʼəβəβəm ‘he shaved’ → kʼəwəwim ‘one shaved’

3.5.2.2 Quadriradicals with total reduplication (1212) Quadriradical verbs with total reduplication have the shape 1212, but they are not derivations of corresponding biradicals or other verbs.⁷⁷ Like the quadriradicals with final reduplication (1233), they often express a notion of repetitive actions or small movements, for example kʸɨfəkʸəf (√kʸfkʸf⁷⁸) ‘sprinkle’, kɨsəkəs (√ksks) ‘dash to bits’ or dɨfədəf (√dfd) ‘mix flour with water’. Also, palatalization and labialization usually affects both parts of the reduplication, illustrated with the Impersonal forms of kɨtəkət (√ktkt) ‘crush s.th. up’ in (150).

⁷⁴ For the verb kʼɨmətʼətʼ ‘wrinkle’ I have once recorded the Jussive base kʼmətʼtʼ with the vowel ə after the second rather than after the first radical (but not in the Imperative). It is not clear whether this was a (performance) mistake or if some verbs can show variation. ⁷⁵ Note that other verbs with doubled final β do not strengthen the penult: kʼəβəβ ‘shave’, xəβəβ ‘encircle’, tʼəβəβ ‘be narrow’. Probably this has to do with the fact that these are all type A verbs, in contrast to jəpəβ ‘dam’ which is a type B verb. ⁷⁶ Cases with only the final radical palatalized or labialized are also attested, but they are rather rare: tɨmezɨʒ ‘you (s) draw out’ (instead of tɨmeʒɨʒ ~ tɨmeʒʒ), kʼəfəfʷim ‘one cut (fingernails)’ (instead of kʼəfʷəfʷim). ⁷⁷ There are a few cases like kʸɨfəkʸəf vs. kʸaffa ‘drizzle’, the first one with total reduplication and the second as a weak triradical without reduplication. ⁷⁸ Or arguably √kIfkIf.

69 (150)  kɨtəkət-ə-m → kʷɨcəkʷəc-i-m ‘one crushed s.th. up’  yɨ-ktəkt → yɨ-kʷcəkʷc-i ‘one crushes s.th. up’  yə-kətkɨt → yə-kʷəckʷɨc-i ‘one should crush s.th. up’

As can be seen in table 25, this type of verb features the same patterns as ordinary type E verbs, including mutation of the penultimate radical (and only the penult) in the Perfective and Imperfective.

Type    E drəttər drə(t)tr dəndr⁷⁹ ‘thicken’

Table 25: Example bases of a quadriradical verb with total reduplication

However, as exemplified in the following table 26, some verbs usually⁸⁰ delete the second radical in the Jussive.

Type    E zfəzəf zfəzf zəzf ‘put to soak’

Table 26: Example bases of quadriradical with total reduplication deleting the second radical in the Jussive

This deletion occurs in case the second radical is a labial (151), or a dorsal followed by a coronal (152) (Banksira 2000: 176ff.):

(151) a.  tʼɨβətʼəβ-ə-m ‘he emptied’⁸¹  yɨ-tʼβətʼɨβ ‘he empties’  tʼətʼɨβ ← (/tʼəβtʼɨβ/) ‘emtpy!’

b.  cʼɨfəcʼəf-ə-m ‘he destroyed a great number’  yɨ-cʼfəcʼɨf ‘he destroys a great number’  cʼəcʼɨf ← (/cʼəfcʼɨf/) ‘destroy a great number!’

c.  tʼɨmətʼəm-ə-m ‘he rolled, twisted’  yɨ-tʼmətʼɨm ‘he rolls, twists’  tʼətʼɨm ← (/tʼəmtʼɨm/) ‘roll!, twist!’

(152) a.  dɨgədəg-ə-m ‘he filled completely’  yɨ-dgədɨg ‘he fills completely’  dədɨg ← (/dəgdɨg/) ‘fill completely!’

⁷⁹ The first r assimilates to the d, ↗ 2.1.3.3. ⁸⁰ Marginally, no deletion or deletion with compensatory gemination also occurs. For instance the Jussive forms of (151) can at times also be ¿tʼəβtʼɨβ ~ ¿tʼəttʼɨβ, ¿cʼəfcʼɨf ~ ¿cʼəccʼɨf, ¿tʼəmtʼɨm ~ ¿tʼəttʼɨm.

70 b.  cʼɨkʼəcʼəkʼ-ə-m ‘he nagged’  yɨ-cʼkʼəcʼɨkʼ ‘he nags’  cʼəcʼɨkʼ ← (/cʼəkʼcʼɨkʼ/) ‘nag!’

In the verb gɨməkəm ‘chip the rims’, for example, the penultimate is mutated in the Jussive even though the template does not require it (153). This devoicing compen- sates the deletion of the preceding nasal, but this happens only if the penultimate is in an environment that does not block mutation; compare for instance with the Jussive dədɨg ‘fill completely, squeeze in’ (152), which has d rather than t in all bases due to the ultimate obstruent g (obstruents except t blocking devoicing, see page 45).

(153)  gɨməkəm-ə-m ‘he chipped the rims’  yɨ-gməkɨm ‘he chips the rims’  gəkɨm ← (/gəmgɨm/) ‘chip the rims!’

3.5.2.3 Verbs with identical first and second radical (113) There is a group of verbs with the shape 113 with an identical first and second radical (or the mutated variant) in all TAM forms, exemplified in table 27.

Type    kʸəkkʸər kʸə(k)kʸr kə(k)kr ‘hold in armpit’

Table 27: Example bases of verb with identical first and second radical

Verbs of this type all have palatalization or labialization of the first two radicals. The ones with palatalization resemble triradicals of type B in that they surface with mutated penultimate radical in all bases and depalatalization in the Jus- sive.⁸² Following Banksira (2000: 73), however, they can be analyzed as totally reduplicated quadriradical verbs (of underlying roots √1I2 or √1U2), which have deleted their second radical in all bases (rather than only in the Jussive, cf. table 26). The verb kʸəkkʸər ‘hold in armpit’, for example, is clearly connected to the noun kʸɨrkʸɨr ‘armpit’. It is safe to assume that both words are based on a reduplicated root ending in r, which is deleted in the verb but not in the noun. Other examples are gʸəkʸət ‘accompany on departure’ and gʸəkkʸər ‘straighten out, arrange’. The latter is related to the not reduplicated gʸəra ‘naïve’ and the reduplicated gʸɨrgʸr bar ‘blaze’, even though here the semantic connection is not obvious.

⁸² Also the verb ʒəʃʃə ‘be cold, wet’ could be classified as having two identical initial radicals z, but it does not depalatalize in the Jussive like a type B verb. According to Leslau (1979c: 724) “the root is zzy possibly going back to zyzy”. Indeed, at first sight it is not obvious whether the root of this verb is √zzI (i.e. verb type A2) or √zIzI (verb type E), but in my opinion it is more plausible to assume the quadriradical root. Firstly, the Imperfective is yɨ-ʒəʃ with mutation of the final consonant which is a feature of the Imperfective of type E verbs.

71 Consider also the synonymous pair of verbs a-gʷəkkʷər and a-n-gʷɨrəkkʷər⁸³ ‘roar’ (with additional prefixes, ↗ 3.6.2 and ↗ 3.5.3.4), which are both based on the same reduplicated root, but the former one with a deleted first r. Another indication in favor of the reduplicative nature of this class of verbs is the fact that the identical radicals are both palatalized/depalatalized as well as labialized and not only one of them. Consider, for instance, the labialization of both radicals triggered by the 3sm suffix -n (↗ 3.12.1) in the Imperative gʷəkʷt-ɨn ‘accompany him on departure!’ (without suffix gəkt). Finally, assuming that these verbs are quadriradicals, the penultimate is not expected to occur in the mutated form in the Jussive. Yet we do not have forms like *gəgt (<*gətgt) but gəkt ‘accompany on departure’ instead with devoicing compensating for the deletion of the preceding radical.

3.5.2.4 Frequentative (medial reduplication 1223) The so-called Frequentative, a largely productive derivation, is formed by reduplication of the medial radical of triradical verbs and inserting either the vowel ə or a between the two reduplicated radicals (1ə2ə3 → 12ə2ə3 / 12a2ə3). Frequentatives also exist for derived stems but often with the vowel a only (for example t-1ə2ə3 → t-12a2ə3). Consider (154) exemplifying medial reduplication by the Perfective of a simple and a derived stem. As mentioned above, refer to section 3.10 for a detailed discussion of the forms and functions of the Frequentative.

(154) a. 1ə2ə3 → 12ə2ə3 / 12a2ə3 √sβr səppər → sɨβəppər / sɨβappər ‘break’ → ‘break frequently’

b. t-1ə2ə3 → t-12a2ə3 √kʼβr təkʼʸəppər → təkʼʸβappər ‘receive’ → ‘receive from each other’

3.5.3 Special verb types and irregular verbs 3.5.3.1 Type 1rD Triradical verbs with an ultimate dental plosive (d, t, tʼ) and a mutated penultimate r→n in the Perfective drop the vowel ə between these two radicals (table 28). Thus we have, for example, wəndəm ‘he went down’ (*wənədəm) or məntʼəm ‘he peeled’ (*mənətʼəm). The Imperfective and Jussive do not have a nasal and behave like normal type A1 verbs.

⁸³ Also common is a-n-gʷɨrakkʷər with -a-.

72 Type   

A1 fənd fərd frd ‘judge’

Table 28: Example bases of verb with penultimate r and ultimate dental

Note that the nasal-dental consonant cluster is never split. The epenthetic vowel that becomes necessary with subject markers beginning with a consonant always follows the dental, for example antʼɨxum ‘you (pm) cut’ (*anɨtʼxum) or wəndɨxʷɨm ‘I went down’ (*wənɨdxʷɨm).

(155) Verbs of type A1-1rD antʼ √Artʼ ‘cut’ fənd √frd ‘judge’ fəntʼ √frtʼ ‘split’ kʼəntʼ √kʼrtʼ ‘despise’ məntʼ √mrtʼ ‘peel’ wənd √Urd ‘go down’

3.5.3.2 Type 1rAI There is a small group of verbs with an unusual shape. As exemplified in table 29, there is no vowel (except epenthetic ɨ) after the first radical in the Perfective and Imperfective and a palatalized r→y as second consonant; the depalatalized r appears, for example, in the Perfective 3pm kʼɨraβom ‘they waited’ or the Infinitive kʼərəβot.

Type    kʼyə / kʼra kʼyə / kʼrə kʼəyə / kʼərə ‘wait’

Table 29: Example bases of verb with the root 1rAI

(156) Verbs of type 1rAI cʼɨyə √cʼrAI ‘stink’ gɨyə √grAI ‘be bored’ kʼɨyə √kʼrAI ‘wait’ sɨyə √srAI ‘buy’ tɨyə √trAI ‘be visible’ wɨyə √UrAI ‘go down’

It is not entirely clear what root these verbs have. Leslau (1979c: liv) states that at least some of them go back to √1rħ, with the loss of the laryngeal ħ, but

73 this does not explain the palatalization of r→y, which occurs also in most other Gurage varieties, sometimes not as palatalization of r but for instance as kʼʸ in Muher kʼʸɨrə or e in Wolane kʼerə. Judging from the vowels in Gumer, these verbs seem to be quadriradicals underlyingly. Like quadriradicals of type E they have no vowel after the first and second radical in the Perfective and Imperfective, but an ə in the Jussive. The second radical r consequently is not the penultimate, which explains that it is not mutated. The supposed loss of a laryngeal ħ leads to the assumption of the vocoid A as third radical, which is the reflex of former laryngeals also elsewhere. This is supported by the presence of a vowel a in the third persons Perfective as in kʼɨraβom ‘they waited’ (↗ 3.16). Finally, one might adopt a final radical I which causes the palatalization. This scenario parallels Im- peratives of the feminine singular, which is formed by palatalization (↗ 3.11.4), in triradicals with a final A. Compare an assumed kʼɨyə < *kʼrAI with bɨyə 2sf. < bɨra+I (2sm. + ) ‘eatǃ’. Nevertheless, it must be left open whether one or all verbs of this type are really based on a root √1rAI and whether such an analysis reflects any diachronic development. Consider, however, Lowenstamm (1996) who reaches the same conclusion. At first sight, the verb nɨyə ‘sleep’ seems to belong to the same group, but it behaves differently. In particular the initial n is geminated in the Imperfective. Table 30 shows its bases.

Type    - nyə / na nnyə / nnrə nay / na ‘sleep’

Table 30: Bases of the verb nɨyə ‘sleep’

3.5.3.3 Verbs with penultimate r and final I or U (1rI / 1rU) Verbs featuring final I or U and a penultimate radical r appear with only one consonant in bases that do not require mutation, as for example in the Imperfective and Jussive of verb type A (table 31). The final weak radical I palatalizes r to y, which then merges with the preceding vowel. Consider, for instance, the type A verb gənə √grI ‘cast a spell’ and its Imperfective ge (gərI → gəy → ge) and Jus- sive gi (gɨrI → gɨy → gi). In bases with mutation r → n no palatalization can take place with the result that the forms look ‘regular’, as for example the Imperfective cənI → cən of the type B verb cənə √trI ‘come’.

Type    A kʼənə kʼe kʼi ‘vanish, be lost’

Table 31: Example bases of verb with penultimate r and ultimate I

74 (157) Verbs with root 1rI gənə √grI ‘cast a spell’ kʼənə √kʼrI ‘vanish’ tənə √trI ‘swear (an oath)’ xənə √xrI ‘dig a hole’ a-mənə √mrI ‘do’

(158) Verbs with root 1rU kʼʷənə √kʼrU ‘roast’ a-fʷənə⁸⁴ √frU ‘rest’

3.5.3.4 Verbs with prefixed n- There are a few verbs that have an additional nasal n- (word initially ɨn-) prefixed to the actual radicals, which often feature a reduplication pattern (table 32). These verbs are not derivations of verbs without n- but simply exist in this form on their own. Apart from the n-, their bases do not differ from equivalent verbs without n-.

Type    nkʼrəkkʼər nkʼrə(k)kʼr nkʼərkʼr ‘move ()’

Table 32: Example bases of verb with prefixed n-

(159) Verbs with prefixed n- ɨngɨβəkəβ √gβgβ ‘thunder, rumble’ ɨnkɨβannər √kβrr ‘roll ()’ ɨnkʼɨrəkkʼər √kʼrkʼr ‘move ()’ ɨnkʼʸəf √kʼIf ‘embrace, include’

3.5.3.5 The verb tʼəppʷə ‘suck’ The verb tʼəppʷə √tʼβU ‘suck’ features the unusual vowels o and u in some forms of the Imperfective and Jussive. They come about because the non-geminated β is labialized by the final radical U, i.e. *βʷ→w. In the persons that do not have a vocalic subject suffix this base-final w fuses then with the preceding vowels ə and epenthetic ɨ (160). When there is a number-gender suffix like -o, the w becomes the syllable onset and does not change (161).

⁸⁴ According to Leslau (1979c: 232), in Chaha and Ezha there is also a verb fʷənə ‘get well, heal’, but I could not confirm this.

75 (160)  *yɨtʼəβʷ → *yɨtʼəw → yɨtʼo ‘he sucks’  *yətʼɨβʷ → *yətʼɨw → yətʼu ‘let him suck’

(161)  *yɨtʼəβʷo → yɨtʼəwo ‘they (m) suck’  *yətʼβʷo → yətʼwo ‘let them (m) suck’

3.5.3.6 The verb bar ‘say’ The very frequent verb bar √βAr ‘say’ shows some irregularities. In the Perfec- tive it loses the base-final r before subject markers beginning with x as shown in table 33. Further, as already mentioned on page 59, the Jussive base features the uncommon vowel ə. See section 3.17 for a lineout of its various occurrences.

  1 ba-xʷ-ɨm ban-nə-m 2m ba-xə-m ba-xu-m 2f ba-xʸ-ɨm ba-xma-m 3m bar-ə-m bar-o-m 3f bar-əc-ɨm bar-əma-m  bʷar(-i)-m

Table 33:  of bar ‘say’

3.5.3.7 The verbs aβ ‘give’ and od ‘tell’ The two verbs aβ √Aβ ‘give’ and od √Ud ‘tell’ have only one consonantal radical (the final one) and begin with a vowel (table 34). They can be viewed as consisting of radical 2 and 3 only where radical 2 is weak.

Root    √Aβ aβ (i)β aβ ‘give’ √Ud od ud od ‘tell’

Table 34: Bases of the verbs aβ ‘give’ and od ‘tell’

The verb aβ ‘give’ goes back to the Semitic root √whb (for example Arabic wahaba ‘give’, Leslau 1979c: 641). In East Gurage languages like Wolane (Meyer 2006: 94) it has the (expected) form wāβ with the vowel a in place of the laryngeal h. In Western Gurage languages, however, the initial w eroded,⁸⁵ resulting in a verb with the reduced root √Aβ. Apart from that, it basically behaves like a verb of type A-1A3-Ø, which have no vowel in the Imperfective and, if possible,

⁸⁵ The related noun waβi ‘generous’ still shows the initial w.

76 a palatalized first radical. Accordingly, the Imperfective base of aβ consists only of the consonant β and shows the vowel i as a remnant of the palatalization. Due to phonological reasons the subject marker y- absorbs the i and the base consists only of β (162).

(162) a. t-iβ vs. b. yɨ-β 3sfS-give.IPFV 3smS-give.IPFV ‘she gives’ ‘he gives’

The verb od ‘tell’ seems to occur in all Gurage varieties and other Ethiosemitic languages (for example Silt’e ēwəd, Endegeñ ēd, Muher ɔd; Gafat əwəd ‘speak’), but quite likely it is a loan from a Cushitic language Leslau (1979c: 112). It is therefore not clear if it makes any sense to postulate a root, and if yes it is not evident what it should be. According to Wetter (2010: 106), the root of the corresponding verb awid ‘tell’ in Argobba is √wr (i.e. √Ur), but as he also mentions this does not account for the initial a, which resembles (but is not) the causative derivation (↗ 3.6.2). Also in Gumer the vowel o of od could be explained as a fusion of a+u at first sight. On the other hand, however, the Imperfective base is ud with u, a vowel that cannot be assumed to contain a. One possible scenario is illustrated in (163): the verb has the root √Ud and the first radical is missing. If put into the template of verb type A2, the forms with the sequence /Uə/ yield the vowel o while /U/ without following ə is realized as u. As already discussed in section 2.2, the vowel o can be analyzed as biphonemic consisting of /ə/+/U/. In the case of od the elements occur in the reverse order /U/+/ə/ (which usually would result in wə as it is for example the case with wər, ↗ 3.5.3.8), but at least to some extent it is still comparable with the decomposition of o to a labialized consonant followed by ə (i.e. Cʷə) (↗ 2.2). Nevertheless, all in all the forms of od are probably best regarded as irregular.

(163) template phonemic base verb form  (1ə)2ə3- → /Uəd-/ → od- → od-ə-m ‘he told’  -(1ə)23 → /-Ud/ → -ud → y-ud ‘he tells’  -(1)2ə3 → /-Uəd/ → -od → y-od ‘let him tell’

3.5.3.8 The verb wər ‘go’ The bases of the verb wər √Ur ‘go’ are formed irregularly (table 35). The verb goes back to a Semitic root √ħwr, realized, for example, as ħorə in Ge’ez or ħāra in Harari (Leslau 1979c: 660). In Gumer, the initial laryngeal is lost reducing the root to √Ur. The Imperfective base, however, unpredictably drops the w and features the vowel a instead, which elsewhere is the remnant of a lost laryngeal (as for example anəβ ‘milk’ √Arβ < √ħlb or bəna ‘eat’ √βrA < √blʕ). Based on the Imperfective, the root would be rather √Ar, but taking the Perfective as the more basic form, √Ur is in preference.

77 Root    √Ur wər ar wər ‘go’

Table 35: Bases of the verb wər ‘go’

3.5.3.9 Exceptionally geminating verbs The verbs fennəkʼ ‘spit out milk’, ənnəkʼ (~ ennəkʼ) ‘vomit’ and əkkəs ‘wait’ are unusual (table 36). Like type B verbs they feature mutation in all bases, but exceptionally the penultimate radical is geminated throughout even though they do not have a weak final radical or r as the expected trigger for gemination (cf. page 45). A fourth verb that seems to fit into this group is əkkʸər⁸⁶ ‘plane wood’. While it predictably geminates before final r, it unexpectedly does not reduce its gemination in the Imperfective and Jussive when followed by an additional morpheme (↗ 3.3.2). Furthermore, it does not depalatalize in the Jussive, and neither does fennəkʼ (i.e. *fənnkʼ).

Root    √frkʼ (?) fennəkʼ fennkʼ fennkʼ ‘spit out milk (child)’ √Irkʼ (?) ənnəkʼ nnkʼ (ə)nnkʼ ‘vomit’ √Iks (?) əkkəs kks (ə)kks ‘wait’ √Ikr (?) əkkʸər kkʸr (ə)kkʸr ‘plain wood’

Table 36: Bases of geminating verbs

As for the root of the verbs with initial ə, one might assume that they have an initial weak radical I. In favor of this analysis is the fact that ənnəkʼ has a variant ennəkʼ with fronted vowel ə → e, and similarly əkkʸər has a fronted initial vowel in other Gurage varieties, for example īnkəl (Zay), ekkər (Endegeñ) and ekʸər (Inor and Gyeto) (Leslau 1979c: 33). A weak initial radical as opposed to the palatalizing element  of type B verbs could also hold as explanation that there is no depalatalization in the Jussive.⁸⁷ Thus, if these verbs are not of type B, the consistent gemination might be compensatory for the initial weak (or ‘missing’) radical. This is in accordance with weak verbs featuring initial e (↗ 3.5.3.10) that also geminate in all bases. Furthermore, also the doubly weak verb anna √ArA ‘defecate’ geminates in the Perfective, whereas n of verbs with of the structure √CrV remains a singleton (for example bəna √βrA ‘eat’, cənə √trI ‘come’). Finally, the question why əkkəs does not have palatalized kʸ, whereas əkkʸər does, can be explained by the fact that the former is a Cushitic loan (Sidamo kēʃʃi, Leslau

⁸⁶ Nevertheless, the data are not complete and it seems that reduction of gemination and depalatalization is also possible, but this has to be checked. ⁸⁷ However, what is then the root of fennəkʼ, which also does not depalatalize, if not √frkʼ?

78 1979c: 34). However, in this case the ə seems to be prothetic similar to the verbs in section 3.5.3.11.

3.5.3.10 Verbs with initial e As indicated on page 58, there are some verbs with initial e. Due to the fact that there is no verb beginning with y (in contrast to initial a and w) it would not be far-fetched to attribute these verbs a first radical I. Furthermore, the realization of initial yə as e is also found in the negative verb of existence enə (cf. table 81) which corresponds to Amharic yəllə ’there is notʼ. As table 37 shows, the two attested verbs ella ‘covet’ and effə ‘cover with lid’ geminate in all bases (comparable to other verbs with initial weak radical). The verb ella is very unusual for its geminated l, but note again that it is a Cushitic loan, for example Kʼabeena hele’a- ~ hela-(Leslau 1979c: 38) or Libido heela’- (Joachim Crass, p.c.).

Root    √IlA ella ella ella ‘covet, desire’ √IfI effə iff (~eff)⁸⁸ eff ‘cover with lid’

Table 37: Bases of verbs with initial e

3.5.3.11 Verbs with additional initial ə There are a few verbs that have an additional vowel ə preceding the first radical, ərəkkʸə ‘throw’ and əʃəʃʃə ‘massage’ being two (slightly distinct) examples (table 38). The verb ərəkkʸə is probably related to Amharic ləgga √lgA ‘hit and throw a ball’ with regular change l → r (Leslau 1979c: 88). The initial ə therefore seems to be purely prothetic, a phenomenon that occurs sometimes in combination with r (cf. Amharic lam vs. Gumer əram ‘cow’, or Amharic rob vs. Gumer əro ‘Wednes- day’).⁸⁹ According to Leslau (1979c: 101) the verb əʃəʃʃə ‘massage’ is a reduplicated stem of *aʃə (< *√ħsy) connected to the Semitic root √ħss ‘stroke; rub’. Thus, the initial ə could be explained as remnant of the a.

Root    √rgI ərəkkʸə rəkʸ (ə)rəgʸ ‘throw’ √sIsI əʃəʃʃə ʃəʃ (ə)ʃəʃ ‘massage, rub’

Table 38: Bases of verbs with prothetic ə

⁸⁸ It is not clear which variant is more acceptable, but there seems to be some variation (as well as in the Jussive). ⁸⁹ Presumably the function of prothetic ə is to preserve the sound r, which otherwise would change to n word-initially (↗ 2.1.3.1).

79 3.5.3.12 Amharic loan verbs Since it is easily possible to borrow words and constructions from the related and structurally similar Amharic, there are also many Amharic verbs used in Gumer. While the subject conjugation is adapted to the Gumer paradigms, the verb bases are usually taken over as they are. Accordingly, gemination is retained, there is no mutation, and unusual consonants appear as for example l, geminated ww or yy, or “non-geminated” n. Table 39 lists four examples, all of them type B verbs borrowed from Amharic, followed by sentences they are attested in. Note that while the first two are common in the daily language, the last two represent rather spontaneous borrowings.

Root    √dwl dəwwəl dəwwl dəwwl ‘call, phone’ √wsn wəssən wəssn wəssn ‘decide’ √kʼys kʼəyyəs kʼəyys kʼəyys ‘plan, put up a scheme’ √lbd ləbbəd ləbbd ləbbd ‘glue on, cover’

Table 39: Example bases of some verbs borrowed from Amharic

(164) dəwwəl-ə-ni-m. phoneѦ.PFV-3smS-BEN.1s-M ‘He called me.’ (165) at-əta bə-cəxa yɨ-rəβɨr-xəma wəssən-o-m. one-3sPOSS LOC-Chaha 3smS-live.IPFV-COMP decideѦ.PFV-3pmS-M ‘They decided that one of them live(s) in Chaha.’ (166) səβat-m-əxʷna gojjo kʼəyyəs-o-m. seven-ALSO-3pmPOSS houseѦ planѦ.PFV-3pmS-M ‘The seven of them planned houses.’ (167) gɨyə-m fərəz-ɨm y-anə-wə sɨkʼɨl⁹⁰acənə-c-ɨm dog-ALSO horse-ALSO REL-EX-MAL.3sm picture bring.PFV-3sfS-CV.M ləbbəd-əc-ɨm. glue.onѦ.PFV-3sfS-M ‘She brought a picture with a dog and a horse on it and glued it [on the wall].’

Sometimes there are ‘minimal pairs’ of an original and a borrowed verb that contrast only in their gemination (i.e. in the Perfective) as kʼənəs ‘begin’ and kʼən- nəs ‘decrease’ (168). Note, however, that there are also the native Gumer verbs kʼəppər ‘decrease’ and anəs ‘be(come) less’.

⁹⁰ Arguably this a loan word from Amharic sɨʼɨl (ሥዕል) ‘picture, painting’, where the ejective kʼ is the rendering of the glottal stop, which is not part of the consonant inventory of Gumer.

80 (168) ‘begin’ ‘decrease’ (< Amharic)  kʼənəs-ə-m kʼənnəs-ə-m  yɨ-kʼərs yɨ-kʼənnɨs  yə-kʼɨrs yə-kʼənnɨs

3.5.3.13 Suppletive Imperatives The verbs cənə ‘come’ and acənə ‘bring’ have suppletive Imperatives both based on an element beginning with n- (table 40). Their suffixes, however, are from different sets: while namə ‘bring; give’ features the common Imperative suffixes -Ø and + in the singular and -o and -əma in the plural (↗ 3.11.3), ne- ‘come’ takes the Perfective subject markers (↗ 3.11.1). Likewise, the latter suffixes also occur on the third special Imperative with yə- meaning ‘take, here you are’.

‘bring; give’ ‘come’ ‘take, here you are’ 2sm namə ne-xə yə-xə 2sf nem ne-xʸ yə-xʸ 2pm nam-o ne-xu yə-xu 2pf namə-ma ne-xma yə-xma

Table 40: Suppletive imperatives

The suppletive Imperative for ‘take’ is exclusively used when handing over something to someone. It corresponds therefore rather to ‘here you areǃ’ than to ‘takeǃ’, a fact that is easily comprehensible given that it is based on yə- ‘ (for, to)’ (Leslau 1979c: 686). The verbs of taking such as tʼəβətʼ, bətta or wəsəd all have their regular Imperatives which mean ‘take from somewhere or someone’ rather than ‘take/accept from me, here you are’. Similarly, the suppletive Imper- ative namə ‘bring, give’, as a manner of speaking the opposite of yə-xə, is used when one wants someone to give or bring something to oneself. The regular Im- perative of ‘give’, aβǃ, is a command to give something to somebody else. Compare the two different Imperatives of ‘give’ in (169) and (170) and their different object markers, i.e. 1st vs. 3rd person. Further, also consider the 1st person benefactive with namə (171) which is the Imperative used to say ‘bring me something (from somewhere), (go and) get me something’ in contrast to namə with the primary object meaning ‘give me, hand over to me’. Note that namə also occurs without object suffix (172).

(169) ɨxa nam-eǃ water bring.IMP[.2smS]-1sO ‘Give me waterǃ’ (170) ɨxa aw-ɨnǃ water give.IMP[.2smS]-3smO ‘Give him waterǃ’

81 (171) ɨxa namə-niǃ water bring.IMP[.2smS]-BEN.1s ‘Bring me waterǃ’ (172) namə nɨ-wra-n, namə nɨ-wra-n! give.IMP[.2smS] 1sS-eat.JUS-3smO give.IMP[.2smS] 1sS-eat.JUS-3smO ‘Give [me and] let me eat it, give [me and] let me eat itǃ’

3.6 Derived stems There are three (mainly) valency-changing prefixes: t(ə)-, a-, and at-. Directly preceding the ‘bare’ verbal roots they form the so-called derived stems. Their templates, however, differ in some cases from the plain stems without prefixes. Table 41 is an overview of the templates of the attested derived stems of sound triradical verbs. Examples of derived stems of weak verbs as well as quadriradicals are found in the corresponding sections.

Prefix Type    t(ə)- A t(ə)-1ə2ə3 t-1ə2ə3 t(ə)-1ə2ə3⁹¹ -a- t(ə)-1a2ə3 t-1a2ə3 t(ə)-1a2ə3 / t(ə)-1a2ə3 B t(ə)-1ə2ə3 + t-1ə2ə3 + t(ə)-1ə2ə3 – -a- t(ə)-1a2ə3 + t-1a2ə3 + t(ə)-1a2ə3 + / t(ə)-1a2ə3 + C t(ə)-1a2ə3 t-1a2ə3 t(ə)-1a2ə3 a- A a-1ə2ə3 a-1ə23 a-123 -a- a-1a2ə3 a-1a23 a-1a23 B a-1ə2ə3 + a-1ə23 + a-1ə23 – C a-1a2ə3 a-1a23 a-1a23 D a-1ʷə2ə3 a-1ʷə23 a-1ʷə23 -a- a-1ʷa2ə3 a-1ʷa23 a-1ʷa23 at- A at-1ə2ə3 at-1ə23 at-1ə23 -a- at-1a2ə3 at-1a23 at-1a23 B at-1ə2ə3 + at-1ə23 + at-1ə23 – -a- at-1a2ə3 + at-1a23 + at-1a23 + C at-1a2ə3 at-1a23 at-1a23

Table 41: Templates of derived stems of sound triradical verbs

For each prefix there are two possibilities, a ‘neutral’ one with the (default) vowel ə⁹² and one with the (additionally infixed) vowel -a- between the first two radicals.

⁹¹ Jussive t(ə)-1ə2ə3 with mutation is also attested, for example təxəttər ‘dressǃ’ (normally təxədər), but these forms are very rare in such a way that they are probably occasional performance mistakes or idiolectal variantions rather than systematic formations. ⁹² In the case of a- of type A there is no vowel in the Jussive.

82 Forms with -a- usually express the notion of reciprocity (↗ 3.7.1.1). Note that type C verbs, which already have a vowel a in their underived form, do not occur with ə. There is probably no verb that appears in all possible six combinations of the three prefixes t(ə)-, a-, and at- and the two vowels ə and a. For example, while some forms are common, only two verbs could be found for a derived stem with a- plus inserted -a-⁹³ and no equivalent form for a type B verb.

3.6.1 Derived stems with t(ə)- (detransitivizer) The prefix t(ə)- can be attached to all verb types⁹⁴ but the templates undergo some changes (table 42). With verbs of type A the template differs from the basic verb in the following two points: all TAM forms feature twice the vowel ə (i.e. one between first and second and one between second and third radical) and there is mutation of the penultimate radical not only in the Perfective but in the Imper- fective as well. Similarly, also verbs of type B show twice the vowel ə in all TAM forms rather than only the first one in the Imperfective and Jussive. In other respects they feature mutation in all TAM forms and depalatalization in the Jussive as they do in the corresponding basic forms. Verbs of type C keep their characteristic vowel a after the first radical, but other than the basic forms there is the vowel ə between second and third radical in all TAM forms. Finally, the few examples of type C all showed mutation in the Jussive.

Type    A t(ə)-xəttər t-xəttər t(ə)-xədər ‘dress (onesel)’ B t(ə)-ʒəppər t-ʒəppər t(ə)-zəppər ‘return’ C t(ə)-manəx t-manər t(ə)-manəx ‘be taken prisoner’

Table 42: Example bases of t(ə)-

(173) Verbs of type A + t(ə)- tədəməd √dmd ‘gather’ təgəttər √gdr ‘lie down’ tərəməd √rmd ‘be expensive, dear’ tərəkəβ √rxβ ‘be found’ təsəppər √sβr ‘break ()’ təxəttər √xdr ‘dress’ tagəd √Agd ‘be tied, be taken prisoner’ tatʼəβ √Atʼβ ‘wash onesel’

⁹³ The two attested forms are type A verb arakəβ ‘announce search, inquire as to the whereabouts of lost cattle or goods’ and type D verb afʷagəg ‘cut off flesh from bone’ (the latter seemingly not occuring in the underived basic stem). ⁹⁴ No type D is attested in my data.

83 təfətta √ftA ‘be loosened, be untied’ tərəssa √rsA ‘get up; forget’ təsəmma √smA ‘feel, sense’ təkʼetʼ √kʼItʼ ‘be traded’ təʃem √ʃIm ‘hide onesel’ təkʼʸəmmə √kʼmI ‘lose, be defeated’ təməccə √mtI ‘be convenient’ təβer √βAr ‘be said’ təmar √mAr ‘study’ tərakʼ √rAkʼ ‘grow up (child)’ təsar √sAr ‘ask’ təʃad √ʃAd ‘take a share’ təxʸər √xAr ‘be known’ tətos √tUs ‘be in a mess’

(174) Verb of type B + t(ə)- təfəkkʼʸər √fkʼr ‘play, sing’ təkʼʸəppər √kʼβr ‘receive, accept, welcome’ təkʼʸənəβ √kʼrβ ‘be near’ təkʼʸəttʼər √kʼtʼr ‘ride pillion’ təʃəkət √skt ‘be fixed, become well behaved’ təxʸəttər √xtr ‘follow’ təʒəppər √zβr ‘return ()’ təcʼənə √tʼrI ‘be born’

(175) Verb of type C + t(ə)- təβannər √βrr ‘be demolished, fall apart’ təmanəx √mrx ‘be taken prisoner’

The prefix t(ə)- together with the vowel -a- between the first and second radical conveys the meaning of reciprocity (table 43) (↗ 3.7.1.1). Except for -a-, there is no change in the template of type A verbs with t(ə)- and (default) vowel ə. Minor exceptions are verbs of type B with t(ə)- and -a- which do not depalatalize in the Jussive, and there are verbs that show mutation in the Jussive and others that do not. Verbs of type C, already possessing a vowel a after the first radical, cannot form an additional reciprocal form.

84 Type    A t(ə)-rakəβ t-rakəβ t(ə)-raxəβ ‘meet (each other)’ B t(ə)-jakəm t-jakəm t(ə)-jakəm ‘bash each other’ ~ t(ə)-jagəm

Table 43: Example bases of t(ə)- + -a-

(176) Verbs of type A + t(ə)- + -a- tədanəg √drg ‘hit each other, fight’ təragəd √rgd ‘touch each other’ tərakəs √rks ‘quarrel’ tərakəβ √rxβ ‘meet (each other)’ təsatəβ √sdβ ‘insult (each other)’ tətʼabətʼ √tʼβtʼ ‘get stuck, hold each other’ təwakka √UgA ‘fight (with one another), stab one another’

(177) Verbs of type B + t(ə)- + -a- təjakəm √dgm ‘bash each other’ təjanəg √drg ‘bump into’ təcaffər √tfr ‘give mouthful to each other’ təcʼabəkʼ √tʼβkʼ ‘be pressed, be glued’ təʃamʷəx √smʷx ‘lean on, lean back’⁹⁵

3.6.1.1 Allomorphs tə- vs. t- The two allomorphs tə- and t- alternate depending on their position in the word. The variant with vowel appears word-initially, whereas t- is used when there is another preceding morpheme of any nature. Imperfectives (178) and Jussives (179) have prefixed subject markers and thus always occur with t-. In the prefixless Imperatives (180) the morpheme appears in the word-initial form tə-.

(178) yɨ-tkʼʸəppər 3smS-accept.IPFV ‘he accepts’ (179) yə-tkʼəppər 3smS-accept.JUS ‘let him accept’

⁹⁵ There is no underived stem *ʃəmʷəx of this root. Further note that Leslau (1979c: 579) does not have labialized mʷ, i.e. təʃaməx, but this contradicts my data.

85 (180) təkʼəppər accept.IMP[.2smS] ‘accept!’

Perfectives do not have prefixed subject markers and thus feature word-initial tə-.

(181) təgəttər-ə-m təʒəppər-ə-m təmanəx-ə-m lie.down.PFV-3smS-M return.PFV-3smS-M be.captured.PFV-3smS-M ‘he lay down’ ‘he returned’ ‘he was captured’

However, if a Perfective form receives a prefix such as subordinators (182)-(183) or negation (184)-(185), as well with the Prohibitive that uses the Perfective base (186), the word-internal allomorph t- is used.

(182) səβ bə-tmanəx-ə (< təmanəxə) person COND-be.captured.PFV-3smS ‘if somebody gets captured’ (183) yə-txʸəttər-ə-n zərma (< təxʸəttərə) REL-follow.PFV-3smS-3smO young.man ‘the young men that follow him’ (184) an-tɨrəssa (< tərəssa) NEG-get.up.PFV[.3smS] ‘he did not get up’ (185) gən-ɨnyə an-tɨʒəppər-xə? (< təʒəppərə) country-DIR NEG-return.PFV-2smS ‘Did you not return to the coutryside?’ (186) ɨn-tɨwakkʷeǃ PROHIB-fight.PFV.IPS.3smO ‘One do not fightǃ’

The same is true when the Infinitive with -ot is furnished with negation an- (187) or another prefix (188). Similarly, the prefix wə- of the other Infinitive always features t- (189) (↗ 3.13).

(187) an-tɨfəkkʼər-ot (<  təfəkkʼər) NEG-play-INF ‘not play(ing)’ (188) yə-tfəkkʼər-ot (<  təfəkkʼər) ATTR/DAT-play-INF ‘of/for playing’ (189) wə-tmanəx (<  təmanəx) INF-be.taken.prisoner ‘be taken prisoner’

Finally, if tə- is prefixed to a verb that begins with a vowel, ə is dropped also in word-initial position.

86 (190) tatʼəβ-ə-m (*təatʼəβəm) < atʼəβəm ‘wash ()’ wash.oneself.PFV-3smS-M ‘he washed himsel’ (191) tagəd-ə-m (*təagədəm) < agədəm ‘tie’ be.tied.PFV-3smS-M ‘it was tied’

3.6.1.2 Assimilation of t- When the prefix t- is adjacent to a consonant (i.e. the first root consonant of a verb) the former assimilates to the latter in some cases⁹⁶ (cf. Banksira 2000: 13f.). Firstly, if the first root consonant is d or j, t- is obligatorily voiced.

(192) d  /yɨ-t-dəməd/ → yɨddəməd ‘gather’  /yə-t-dəməd/ → yəddəməd j  /yɨ-t-janəg/ → yɨdjanəg ‘bump into’  /yə-t-janəg/ → yədjanəg There is optional voicing assimilation with the other voiced obstruents (according to Banksira 2000: 13 only in fast speech), i.e. g (193), z and ʒ (see (195) below). Furthermore, t- can optionally be realized as ejective when followed by kʼ (194).

(193) g  /yɨ-t-gəttər/ → yɨtgəttər~ yɨdgəttər ‘lie down’  /yə-t-gədər/ → yətgədər ~yədgədər (194) kʼ  /yɨ-t-kʼaw/ → yɨtkʼaw~ yɨt’kʼaw ‘drink coffee’  /yə-t-kʼaw/ → yətkʼaw ~yət’kʼaw

All alveolar fricatives (s, ʃ, z, ʒ) can occasionally feature total assimilation of t-. Nevertheless, no assimilation (apart from the optional voicing with z and ʒ) is much more common.

(195) s  /yɨ-t-sər/ → yɨtsər~ yɨssər ‘ask’  /yə-t-sar/ → yətsar ~yəssar ʃ  /yɨ-t-ʃem/ → yɨtʃem~ yɨʃʃem ‘hide ()’  /yə-t-ʃem/ → yətʃem~ yəʃʃem z  /yɨ-t-zɨrakʸ/ → yɨtzɨrakʸ ~ yɨdzɨrakʸ ‘talk together’ ~yɨzzɨrakʸ  /yə-t-zangʸ/ → yətzangʸ~ yədzangʸ ~yəzzangʸ

⁹⁶ This is in contrast to Amharic, where the cognate t- completely assimilates to a following consonant in all cases resulting in geminates, for example təkʼəmmətʼə ‘he sat down’ → yɨkkʼəmətʼ ‘he sits down’ or təfəlləgə ‘he was wanted / looked for’ → yɨffəlləg ‘he is wanted / looked for’ (cf. Leslau 1995: 468)

87 ʒ  /yɨtʒɨβappər/ → yɨtʒɨβappər~ yɨdʒɨβappər ‘be inverted; ~yɨʒʒɨβappər return fre-  /yə-t-zangʸ/ → yətzangʸ~ yədzangʸ quently’ ~yəzzangʸ There is no assimilation when the root consonant is a sonorant as in (196).

(196) r  /yɨ-t-rakəβ/ → yɨtrakəβ (*yɨdrakəβ) ‘meet’  /yə-t-raxəβ/ → yətraxəβ (*yədraxəβ) β  /yɨ-t-βeβər/ → yɨtβeβər (*yɨdβeβər) ‘say to  /yə-t-βeβər/ → yətβeβər (*yədβeβər) each other’ m  /yɨ-t-mər/ → yɨtmər (*yɨdmər) ‘learn’  /yə-t-mar/ → yətmar (*yədmar)

As for the voiceless alveolar stops c, cʼ and t’, Banksira (2000: 13) states that they cause total assimilation of t-, as for instance yɨ-t-cafər → yɨccafər. Here, however, the notation with ‘unassimilated’ t- has been chosen (197) (in accordance to (192) /t-j/→dj rather than jj).⁹⁷

(197) c  /yɨ-t-cafər/ → yɨtcafər ‘give a mouthful to  /yə-t-cafər/ → yətcafər each other’

cʼ  /yɨ-t-cʼən/ → yɨtcʼən ‘be born’ tʼ  /yə-t-tʼən/ → yəttʼən

Finally, needless to say that t- does not change when it meets t.

(198) c  /yɨ-t-tos/ → yɨttos ‘be in a mess’  /yə-t-tos/ → yəttos

3.6.2 Derived stems with a- (direct causativizer) The prefix a- can be attached to all verb types without changes in their basic templates (table 44). One minor exception are verbs of type A2: their derived causative Jussive bases do not show the vowel ə that is present in the underived Jussive (see page 49). Thus direct causatives have only one template for type A. Verbs of type B do not alter and feature their characteristic loss of palatalization in the Jussive also in the derived causative. As for verb type C, all attested Jussive bases show mutation, whereas the underived Jussives occur either with or without mutation.

⁹⁷ Note that in principal I do not claim that Banksira’s total assimilation is wrong. Given that he is a native speaker of Chaha, I assume that he must feel some difference between tc and cc. Never- theless, I am not sure if there really is a discernible (acoustic or measurable) difference between them. Maybe Banksira’s analysis is somewhat influenced by the Amharic orthography conven- tions? Note that a Gumer speaker suggested to write unassimilated t in Gumer, for example yɨtcaffər <ይትቻፈር>.

88 Type D, which is rare, shows no changes in the template.

Type    A₍₁₎ a-xəttər a-xədr a-xdr ‘dress ()’ A₍₂₎ a-tʼənəkʼ a-tʼərkʼ a-tʼrkʼ ‘dry ()’ B a-cəffər a-cə(f)fr a-tə(f)fr ‘give a mouthful’ C a-zapət a-zapt a-zapt ‘cause to get lost’ D a-wənəs a-wəns a-wəns ‘make feel lonely’

Table 44: Example bases of a-

(199) Verbs of type A + a- aβəssər √βsr ‘cook ()’ afəttʼər √ftʼr ‘hurry; prepare food’ afəzəz √fzz ‘improve’ agəffər √gfr ‘lower’ aməntʼ √mrtʼ ‘escape’ arədəd √rdd ‘light fire, kindle’ arəttʼər √rtʼr ‘melt ()’ atʼəβətʼ √tʼβtʼ ‘give somebody to hold’ atʼəməd √tʼmd ‘catch (fish), trap’ atʼənəkʼ √tʼrkʼ ‘dry (); frighten’ atʼənnər √tʼrr ‘filter, purify’ awənd √Urd ‘bring down’ axəttər √xdr ‘dress ()’ aβas √βAs ‘make worse’ adəkʼ √dAkʼ ‘make laugh’ arəkʼ √rAkʼ ‘make distant, take far away’ axʸər √xAr ‘become clear’ aβətta √βdA ‘mary off one’s daughter’ aβəna √βrA ‘feed’ afəkka √fkA ‘remove, take out’ afətta √ftA ‘gallop’ agəppa √gβA ‘put in, let in; marry’ agəssa √gsA ‘burp’ akəna √krA ‘ascend ()’ asəna √srA ‘bring, take to’ atʼəffa √tʼfA ‘extinguish’ awəna √UrA ‘put, put down’ awəttʼa √UtʼA ‘take out’ axəna √xrA ‘shout, make noise’

89 akʸəs √kIs ‘joke’ akʼetʼ √kʼItʼ ‘trade’ ateg √tIg ‘ask somebody to do (esp. buy) something’ ater √tIr ‘make slowly, be late’ afʷənə √frU ‘take a rest’ akʷəʃʃə √ksU ‘remove upper layer of əssət’ akʼʷəm √kʼUm ‘put up, erect; stop (); step on something’ amʷəkʼ √mUkʼ ‘warm, heat’ arotʼ √rUtʼ ‘make run’ agənə √grI ‘make problems’ amənə √mrI ‘do’ arəʃʃə √rsI ‘plait, weave’ asəkkʸə √sxI ‘make flee’

(200) Verbs of type B + a- aβetət √βtt ‘widen’ ajəggər √dgr ‘trouble ()’ ajəkkʷər √dgʷr ‘make droop, make wilt’ arəkʸəm √rgm ‘make ride, put somebody on horse’ acəffər √tfr ‘give a mouthful’ axʸəttər √xtr ‘make follow, escort’ acənə √trI ‘bring’

(201) Verbs of type C + a- azapət √zβt ‘cause to lose one’s way’ azakkʼʷər √zkʼʷr ‘make somebody speak too much’ aʒanəg √ʒrg ‘make somebody leave’

(202) Verbs of type D + a- awənəs √βrs ‘make feel lonely’⁹⁸ afʷəkək √fkk ‘crawl’

⁹⁸ The w results from an intervocalic bʷ. Note, however, that in Gumer awənəs and abʷənəs coexist, see example (11).

90 Verbs with prefix a- and additional infix -a- are very rare. The only two examples that could be found are represented in table 45. Note that while arakəβ ‘announce search’⁹⁹ is a derivation of nəkəβ √rxβ ‘find’, there is seemingly no other form belonging to the root √fgg than afʷagəg ‘cut off flesh from bone’.

Type    A a-rakəβ a-raxβ a-raxβ ‘announce search’ D a-fʷagəg a-fʷagg a-fʷagg ‘cut off flesh from bone’

Table 45: Examples of a- + -a-

3.6.3 Derived stems with at- (indirect causativizer) The prefix at- is attested with verb types A, B and C¹⁰⁰ (table 46). When prefixed to type A verbs, at- entails some changes to their templates. Firstly, the Jussive base features the vowel ə between the first two radicals, and secondly the Imperfective base shows mutation. Nevertheless, there are only very few examples and the data are sometimes contradicting (or showing variation). Very strikingly, however, there is a group of type A verbs that acquire a palatal element in combination with at-. Thus, they behave basically like type B verbs with mutation in all bases and depalatalization in the Jussive (see below). Finally, the templates of type C verbs remain the same, the Jussive featuring mutation rather than variation.

Type    A at-səppər at-sə(p)pr at-səβr ‘cause to break’ A→B at-gʸəffər at-gʸə(f)fr at-gə(f)fr ‘cause to release’ B at-ʒəppər at-ʒə(p)pr at-zə(p)pr ‘cause to return’ C at-manəx at-manx at-manx ‘cause to be captured’

Table 46: Examples of at-

(203) Verbs of type A + at- addənəg √drg ‘cause to beat’ atrəgəd √rgd ‘cause to touch’ atsəppər √sβr ‘cause to break’

⁹⁹ Leslau (1979c: 524) lists atrakəβ for this meaning in Chaha, while Banksira (2000: 36) has arakəβ. ¹⁰⁰No example for type D could be found. One example of another verb type that I recorded with at- is atɨnnə ‘put to sleep’ derived from the weak quadriradical nɨyə ‘sleep’. Both basic and derived form behave somewhat irregular (cf. table 30). It is not clear whether there are other quadriradical verbs (type E and F) with at- or not.

91 atmər √mAr ‘teach’ atrəkʼ √rAkʼ ‘bring up, raise (children)’ atrotʼ √rUtʼ ‘force to run’ atxʸər √xAr ‘inform’ atβəna √βrA ‘make eat’ attənə √trI ‘cause to swear, have someone take an oath’

(204) Verbs of type B + at- atkʼʸənəβ √kʼrβ ‘bring near, present’ atʃəkət √skt ‘cause to be fixed, have repaired’ adʒəppər √zβr ‘cause to return’ atcot √tUt ‘cause to work, hire’ adʒor √zUr ‘force to go around’

(205) Verbs of type C + at- atmanəx √mrx ‘cause to be captured’ atkʼattʼər √kʼtʼr ‘assist knotting, make knot’ adzakkʼʷər √zkʼʷr ‘cause to talk too much’ atmaccə √mtI ‘make angry’

As mentioned above, some verbs of type A receive an additional palatal element when prefixed with at-. As can be seen in (206), initial velars are palatalized. The alveolar s in atsenəf ‘scare’, on the other hand, is not altered but the first vowel is raised to e.¹⁰¹ This is in contrast to ‘real’ type B verbs where alveolars are always palatalized (for example ʃəkkər √sgr ‘change’). Verbs beginning with radical A as in the second group in (206) seem to occur regularly with a palatal glide and the vowel a changed to ə. In the case of atyəkək ‘itch’ I have also recorded atekək with raised vowel. Finally, note that also atxʸər ‘inform’ in (203) shows palatalization. However, this verb is a special case in that it features palatalization in all derivations, except for the plain Perfective and Jussive forms which are xar and yə-xar respectively.

(206) Verbs of type A + at- +  atgʸəffər √gfr ‘cause to release, deprive’ atkʸətəf √ktf ‘have something hashed’ atsenəfə √srf ‘scare someone’

¹⁰¹For Chaha, Rose (2007: 413) mentiones the form atsyənəf with a glide, but this form could not be confirmed in Gumer. An alternative form seems to be atsənəf without any palatal element.

92 atyəttər √Adr ‘leave s.th. for next day, make spend the night’ atyəkək √Akk ‘itch’ (~atekək) atyəsəs √Ass ‘make sweep’ atyəʃʃə √AʒI ‘show’

The templates of verbs with prefix at- and additional infix -a- (table 47) are almost the same as the ones without -a-. Again, type A verbs show mutation in the Imperfective but not in the Jussive. Type B verbs do not feature depalatalization in the Jussive as it is the case with t(ə)- + -a-.¹⁰²

Type    A at-rakəβ at-rakβ¹⁰³ at-raxβ ‘introduce’ B at-kʼʸappər at-kʼʸa(p)pr at-kʼʸa(p)pr ‘exchange words’

Table 47: Examples of at- + -a-

(207) Verbs of type A + at- + -a- atragəd √rgd ‘cause to touch one another’ atrakəβ √rxβ ‘introduce’ attakkər √tkr¹⁰⁴ ‘chase away’ atβatta √βdA ‘consult one another’ atrassa √rsA ‘help lift’ atwana √UrA ‘make a plan’ atwand √Urd ‘belittle, humiliate’

(208) Verbs of type B + at- + -a- atkʼʸappər √kʼβr ‘exchange words, make reach (e.g. by shouting)’ atcʼamət √tʼmt ‘think, reflect’

¹⁰²Note that atkʼʸappər is the only attested type B verb with at- + -a-. It is not excluded that additional examples would show some variation, especially in the Jussive, as also for atkʼʸappər one instance of no mutation in the Jussive (at-k’ʸaβr) was recorded. ¹⁰³This verb shows variation in the Imperfective base between (regular) mutation at-rakβ and no mutation at-raxβ. The same is attested with t(ə)-: t-rakβ ~ t-raxβ. ¹⁰⁴According to Leslau (1979c: 34) attakkər is related to *akkər (which does not exist in its underived form), thus the root would be √Akr rather than √tkr.

93 3.6.3.1 Assimilation of at- The dental of at- assimilates in basically the same contexts as t- (↗ 3.6.1.2) to a directly following consonant. Note that at- is adjacent to the first radical in all TAM forms, i.e. other than with t(ə)- this rule also applies in the Perfective. Nevertheless, here only a few instances are attested. They show obligatory voicing before the dental d (209) and optional voicing before the other voiced obstruents gʸ (210) and the sibilants z and ʒ¹⁰⁵ (211).

(209) d  /at-dənəg-ə/ → addənəgə ‘cause to beat’  /y-at-dənɨg/ → yaddənɨg  /y-at-dərg/ → yaddərg (210) g⁽ʸ⁾  /at-gʸəffər-ə/ → atgʸəffərə ~ adgʸəffərə ‘cause to be  /y-at-gʸəffɨr/ → yatgʸəffɨr~ yadgʸəffɨr released, free’  /y-at-gəffɨr/ → yatgəffɨr~ yadgəffɨr (211) z  /at-zakkʼʷər-ə/ → atzakkʼʷərə ~adzakkʼʷərə ‘cause to speak  /y-at-zakkʼʷɨr/ → yatzakkʼʷɨr~ yadzakkʼʷɨr too much’  /y-at-zakkʼʷɨr/ → yatzakkʼʷɨr~ yadzakkʼʷɨr ʒ  /at-ʒəppər-ə/ → atʒəppərə~ adʒəppərə ‘cause to re-  /y-at-ʒəppɨr/ → yatʒəppɨr~ yadʒəppɨr turn; get well’  /y-at-zəppɨr/ → yatzəppɨr~ yadzəppɨr

Note that other than with t-, the occasional total assimilation of the dental in at- to a following sibilant seems to be almost impossible, or at least not very common. Apart from one instance assɨyənəf ‘scare’ (normally atsɨyənəf ~ atsenəf ) I have never recorded forms like *azzakkʼʷər, *aʒʒəppər, *assəppər or *aʃʃəkət.

3.7 Uses of t(ə)-, a- and at- The three derivational verbal prefixes are devices that primarily serve to change the valency of verbs. While t(ə)- is a detransitivizer reducing the valency, both a- and at- are causatives increasing it. Most verbs cannot occur with all three derivations. Rather, depending on their semantics and/or transitivity, they allow one or two of them. Any verb stem, be it underived or derived, can only express one value of valency, i.e. change in valency necessitates derivation. As shown in (212) the basic (underived) stems are either intransitive or transitive.¹⁰⁶ Conversely, derived forms can either represent the intransitive part of a verb pair, i.e. with t(ə)-, or the transitive one, i.e. with a-.

(212)   tə-rəssa ‘get up’ ← nəssa ‘lift’ betət ‘be wide’ → a-βetət ‘widen’

¹⁰⁵Another example verb is atʒor ~ adʒor ‘cause to go round’. ¹⁰⁶Possible exceptions are ambitransitives like verbs of ingestion, cf. Amberber (2000: 313).

94 In addition to these regular cases, there is a considerable number of verbs that do not possess a basic underived stem. Especially among these prefix-necessitating (‘bound’) stems there are verbs where the prefixes do not change the valency as such but rather encode valency or express other nuances (cf. Petros 1994 and Banksira 1999a on Chaha, and Amberber 2000 on Amharic).

3.7.1 Detransitivizer t(ə)-: anticausative(-passive-reflexive) The detransitivizer t(ə)- combines with transitive verbs to derive anticausatives, and marginally passives and reflexives.

(213) səppər ‘break ()’ → təsəppər ‘break (); (be broken (by someone))’ cʼənə ‘give birth’ → təcʼənə ‘be born’ atʼəβ ‘wash’ → tatʼəβ ‘wash onesel’

Derivations with t(ə)- fade out the subject (agent) of the original transitive verb (214a) (but see the reflexives below), while the old object syntactically becomes the new subject (214b).

(214) a. b-afər nəssa-n-ɨm. LOC-land lift.PFV[.3smS]-3smO-M ‘He lifted him from the ground.’ b. b-afər cona-m-ta tərəssa-m. LOC-land sit.PFV[.3smS]-CV.M-LINK get.up.PFV[.3smS]-M ‘He sat on the ground and stood up.’

The most common type of the three valency-reducing derivations mentioned above are anticausatives. Like derived intransitive verbs they feature the logical undergoer as subject but other than passives there is no (external) causer implied (cf. Dixon & Aikhenvald 2000, in particular Amberber 2000 for Amharic). A list of some typical anticausatives are given in (215), followed by a few example sentences (216)-(220).

(215) Examples of anticausative derivations ʒəppər ‘return ()’ → təʒəppər ‘return ()’ gəttər ‘put to sleep’ → təgəttər ‘lie down’ nəssa ‘lift’ → tərəssa ‘get up, set of’ ʃem ‘hide ()’ → təʃem ‘hide ()’ fətta ‘untie’ → təfətta ‘become loose’ tɨʼmətʼəm ‘twist ()’ → tətʼmətʼəm ‘twist ()’ gɨrətəm ‘break in two ()’ → təgrətəm ‘break in two ()’ məntʼ ‘peel’ → təməntʼ ‘peel off (skin)’ bɨraccʼə ‘scatter ()’ → təβraccʼə ‘scatter ()’

95 (216) xɨ təʒəppər-o-m bet-əxʷna gəpp-o-m. DEM return.PFV-3pmS-M house-3pmPOSS enter.PFV-3pmS-M ‘Then they returned home.’ (217) tərəss-o-m-tanə sɨddət wər-o-m. stand.up.PFV-3pmS-CV.M-LINK exile go.PFV-3pmS-M ‘They stood up (set of) and went into exile.’ (218) wədərə b-əgr-əna tətʼmətʼəm-ə-m. rope LOC-foot-1sPOSS twist.PFV-3smS-M ‘A rope twisted around my foot.’ (219) ãfʷɨna-na təməntʼ-ə-m. nose-1sPOSS peel.off.PFV-3smS-M ‘My nose peeled off.’ (220) at kilo yɨxər sɨkkʷar təβraccʼə-m. one kilo approximately sugar scatter.PFV[.3smS]-M ‘Approximately one kilo of sugar scattered.’

In contrast to Amharic, for example, the passive use of t(ə)- is much less frequent in Gumer. The verb təcʼənə ‘be born’ as in (221) is one of the very few examples that could be counted as “real” passive.

(221) yə-tcʼənə-x-wə kərə REL-be.born.PFV-1sS-MAL.3sm day ‘my birthday (lit. the day I was born on)’

Otherwise, passive values are generally expressed by means of the Impersonal (↗ 3.11.5). As the alternative translation ‘be broken (by somenone)’ of təsəppər in (213) above demonstrates, passives with t(ə)- are occasionally in use, but most of the time such constructions appear to be copies from Amharic. Consider the following sentence where the speaker first used the derived təmanəx ‘he was captured’ and then corrected themself to the Impersonal of the underived verb manəxʷi ‘one captured (him)’.

(222) əxʷa yə-tmanəx-ə-m səβ yə-manəxʷ-i səβ now REL-be.captured.PFV-3smS-ALSO person REL-capture.PFV.IPS-3smO person bə-məgəra-w-ʃ yɨ-sre-βa. INSTR-calves-COP.3smS-PRAG 3S-buy.IPFV.IPS.3smO-AUX.PT ‘Now, also captured people used to be bought (back) by calves.’ (lit.: ... it is with calves that one used to buy a person that one took prisoner.)

As for reflexives, there are not many verbs that readily fit the common defini- tion saying that they are constructions “in which subject and object are the same entity” (Payne 1997: 198). In most cases reflexivity is expressed by gəg ‘body’ plus possessive suffix as the object of the verb (with or without object marker yə-, ↗ 4.7.1.2).

96 (223) (yə-)gəg-əna b-ɨxa dən aʃʃə-xʷ-ɨm. DAT-body-1sPOSS LOC-water inside see.PFV-1sS-M ‘I saw myself in the water.’ (224) (yə-)gəg-əna ə-tʼəra. DAT-body-1sPOSS 1sS-hate.IPFV ‘I hate myself.’

The clearest instances of reflexive verbs with t(ə)- are grooming verbs like tatʼəβ ‘wash onesel’, but it is important to note that the object, i.e. the body part that is washed, does not have to be demoted (225b), and even the primary object suffix on the verb referring to the (definite) object is possible (225c).

(225) a. tatʼəβ-xʷ-ɨm. wash.PFV-1sS-M ‘I washed myself.’ b. gʷɨnər-əna tatʼəβ-xʷ-ɨm. head-1sPOSS wash.PFV-1sS-M ‘I washed my head/hair.’ c. gʷɨnər-əna tatʼəβ-xʷ-ɨn-ɨm. head-1sPOSS wash.PFV-1sS-3smO-M ‘I washed my head/hair.’

Thus the derived reflexive form with t(ə)- does not reduce valency as such, but it still expresses reflexivity in the sense that the subject is affected. Interestingly, it seems that reflexivity is only possible when the way of washing is established in the culture. For instance, since traditionally one does not brush one’s teeth (i.e. with a toothbrush, water and toothpaste), there is no verb equivalent to ‘brush teeth’. When somebody wants to refer to this activity, it is only possible with the basic transitive verb atʼəβ ‘wash’ (226), very much as one would use it with any other object that is not a body part (227).

(226) sɨn-əna atʼəβ-xʷ-ɨm. tooth-1sPOSS wash.PFV-1sS-M ‘I brushed (lit. washed) my teeth.’ (227) cʼamma atʼəβ-nə-m. shoe wash.PFV-1pS-M ‘We washed shoes.’

3.7.1.1 Reciprocals As already indicated earlier, the infix -a- added to derived stems conveys the meaning of reciprocity. This concerns in particular the t-stems, but also the at- stems (↗ 3.7.3.1), which express causation of reciprocity (cf. Banksira 2000: 36f.).

97 In contrast, a-stems virtually cannot form reciprocals.¹⁰⁷ In (228) a few illustrative examples of reciprocals derived from their basic verb are listed (including one quadriradical).

(228) √drg dənəg ‘hit’ → tədanəg ‘hit each other’ √rxβ nəkəβ ‘find’ → tərakəβ ‘meet’ (“find each other”) √UgA wəkka ‘stab’ → təwakka ‘fight, stab each other’ √tfr cəffər ‘take a → təcaffər ‘give a mouthful mouthful’ to each other’ √dgm jəkəm ‘bash’ → təjakəm ‘bash each other’ √zrgI zɨrəkkʸə ‘speak, talk’→ təzrakkʸə ‘talk together, converse’

Remember that the t-stem without -a- do not express reciprocity, but simply passive (or reflexive or anticausative) as in the following contrastive pairs.

(229) tərəkəβ ‘be found’ vs. tərakəβ ‘meet’ (“find each other”) tətʼəβətʼ ‘be caught’¹⁰⁸ vs. tətʼabətʼ ‘get stuck; hold each other’

Furthermore, needless to say that the vowel a in derived stems of type C verbs is part of the verb root and therefore does not mark reciprocity.

(230) √mrx manəx ‘take prisoner’ → təmanəx ‘be taken prisoner’ (*‘take each other prisoner’) √mtI maccə ‘be angry’ → atmaccə ‘make s.o. angry’ (*‘make each other angry’)

Since reciprocals involve two agents acting on each other, they normally occur with plural subject markers (231)-(234) (cf. Banksira 2000: 36f.).

(231) sɨddɨscɨm-əxʷna bə-βoks tədanəg-o-m. all.six-3pmPOSS INSTR-boxing hit.each.other.PFV-3pmS-M ‘The six of them boxed each other.’ (232) sost gɨred bə-tifə yɨ-ddanəg-əma. three girls INSTR-slap.in.face 3S-hit.each.other.IPFV-pfS ‘Three girls slap each other in the face.’ (233) xʷet gɨred cənə-ma-m agat t-agat-əxnəma two girls come.PFV-3pfS-CV.M shoulder COM-shoulder-3pfPOSS təragəd-əma-m. touch.each.other.PFV-3pfS-M ‘Two girls came and touched each other’s shoulder.’

¹⁰⁷The reason for this must be connected to the fact that most a-stems are causatives of anticausative verbs, which do not have a (human) agent. One exception I am aware of is azrakkʸə √zrgI ‘engage in conversation, talk together’ derived from zɨrəkkʸə ‘speak, talk’. ¹⁰⁸Probably the verb tətʼəβətʼ as a passive is only a loan construction from Amharic (i.e. təyaz)

98 (234) xɨ za səβat səβ səβacɨm-əxʷna at-at bɨxer DEM DEM seven person all.seven-3pmPOSS one-one nation xər-o-m-tanə yɨ-twakk-o-βa. become.PFV-3pmS-CV.M-LINK 3S-fight.each.other.IPFV-pmS-AUX.PT ‘Then, these seven persons fought each other all seven of them as one nation each.’

With singular subjects as in (235), they express a habitual meaning (cf. Banksira 2000: 37). Similarly in (236), the verb is in the singular because it is intended to express that the many listed subjects generally engage in fighting activities and not that they all fight against each other (even though they also fought against each other).

(235) yɨ-twakka səβ 3smS-fight.each.other.IPFV person ‘someone who fights’ (236) gʸəta-m ɨkkɨm, ɨnor-ɨm ɨkkɨm, ɨndəganə-m ɨkkɨm, mʷəxr-ɨm ɨkkɨm, Gyeto-ALSO just Inor-ALSO just Endegeñ-ALSO just Muher-ALSO just əʒa-m ɨkkɨm yɨ-twakka. Ezha-ALSO just 3smS-fight.each.other.IPFV ‘The Gyeto, the Inor, the Endegeñ, the Muher and the Ezha fight just the same.’

Finally, note that reciprocals are also combinable with the Impersonal:

(237) tərakʷəʃ-i-m, təwakkʷe-m, səβ anəkʼ-ə-m. quarrel.PFV.IPS-3smO-M fight.PFV.IPS.3smO-M person be.finished.PFV-3smS-M ‘They (one) quarreled, they (one) fought, and people died.’ (238) oo, təsanəfʷ-i-m. yes fear.each.other.PFV.IPS-3smO-M ‘Yes, they (one) feared each other.’

3.7.2 Causativizer a- (direct causativization) The typical occurrence of the causativizer a- (239) is with intransitive (unaccusative) verbs that express a state and/or the respective change of state (a), the English translation often being rendered with ‘be’ or ‘become’ plus adjective as in (b). Further, verbs of inherently directed motion combine with a- as well (c). Additionally, a small number of transitive verbs also form derived stems with a-. On the one hand, there are the verbs of ingestion, which can be considered ambitransitives (d), and on the other hand there are a few transitive verbs that occur with a-, but only when the derived stem expresses a transfer of possession (e) (cf. Banksira 1999a).

99 (239) a. nəttʼər ‘melt ()’ → arəttʼər ‘melt ()’ b. tʼənəkʼ ‘be(come) dry’ → atʼənəkʼ ‘dry ()’ c. wənd ‘go down’ → awənd ‘bring down’ d. bəna ‘eat’ → aβəna ‘feed’ e. sɨyə ‘buy’ → asyə ‘sell’

Intransitive verbs that are not unaccusative (i.e. verbs with agentive subjects) normally do not form direct causatives with a-, but with at- (↗ 3.7.3). Exceptions are verbs of (directed) motion (245) and verbs of ingestion and transfer of possession (252).

(240) dənəs ‘sing and dance’ → addənəs ‘make dance’ (*adənəs) zəkkər ‘jump’ → atzəkkər ‘make jump’ (*azəkkər)

Direct causativization with a- introduces an external causer who brings about the state or change of state, or following Payne (1997: 181) who is “directly, in- stantly, and probably physically responsible for the effect”. Syntactically, as shown in (241) the new argument (i.e. the causer) becomes the subject and the old subject appears as the object of the causative construction.

(241) a. kʼawa fəttʼər-ə-m. coffee be.ready.PFV-3smS-M ‘The coffee is prepared.’ b. yunus kʼawa afəttʼər-ə-m. Y. coffee make.ready.PFV-3smS-M ‘Yunus prepared coffee.’

In (242) a selection of typical direct causative derivations from intransitive unaccusative verbs is shown, followed by a few more example sentences (243)-(244).

(242) Examples of direct causative derivations mʷakʼ ‘be(come) warm’ → amʷəkʼ ‘warm ()’ bas ‘be(come) bad’ → aβas ‘make worse’ fəzəz ‘be(come) better’ → afəzəz ‘make better’ dɨrəttər ‘be(come) thick’ → adrəttər ‘make thick’ nɨwəʃʃə ‘be(come) familiar, → arwəʃʃə ‘make familiar, learn’ teach’ fəttʼər ‘be(come) ready → afəttʼər ‘make ready, (food)’ prepare (food)’ tʼənnər ‘be(come) filteredʼ → atʼənnər ‘filter’ bəssər ‘cook ()’ → aβəssər ‘cook ()’ nədəd ‘burn ()’ → arədəd ‘set fire, ignite’ jəkkʷər ‘droop, wilt’ → ajəkkʷər ‘make droop’ teg ‘be done → ateg ‘accomplish successfully’ successfully’ zapət ‘get lost’ → azapət ‘make get lost’

100 (243) a. mɨʃt-xʸɨta dakʼ-əc-ɨm. woman-DEF.sf-COP.3smS laugh.PFV-3sfS-M ‘The woman laughed.’ b. nɨkkʼar-u y-adəkʼ-ə-ndə. a.lot-COP.3smS REL-make.laugh.PFV-3smS-1pO ‘[This] made us laugh  .’ (244) a. yə-rədəd-ə bet-xʷɨt ʃərətβet-u. REL-burn.PFV-3smS house-DEF.sm restaurant-COP.3smS ‘The house that burnt down is a restaurant.’ b. y-arədəd-ə kʼar bə-βəna REL-burn.PFV-3smS THING COND-eat.PFV[.3smS] ‘if he eats something that burns (i.e. spicy)’

With verbs of directed motion (‘go’), as well as cənə ‘come’ and səna ‘arrive’, derivation with a- expresses that an object is taken, brought, put, etc. in the respective direction as illustrated in (245) and the following example sentences (246)- (251).

(245) Examples of direct causatives of verbs with directed motion gəppa ‘go in’ → agəppa ‘put in, bring in, etc.’ wəttʼa ‘go out’ → awəttʼa ‘take out, bring out, etc.’ wənd ‘go down’ → awənd ‘take down, bring down, etc.’ kəna ‘go up’ → akəna ‘take up, bring up, etc.’ cənə ‘come’ → acənə ‘bring’ səna ‘arrive, reach’ → asəna ‘bring to, take to’ (246) tʼəwətʼ-ə-n-ɨm-tanə xɨ bəryə agəppʷa-n-ɨm. take.PFV-3smS-3smO-CV.M-LINK DEM B. bring.in[.3smS]-3smO-M ‘He grabbed him and then he brought (entered) him to Berye.’ (247) ɨmmat kʷɨtara-ta awəttʼa-xʷ-ɨn-ɨm. only chicken-3smPOSS take.out.PFV-1s-3smO-M ‘I only took out the chicken.’ (248) mʷɨz-xʷɨta tə-səmɛ awənd-o-m. banana-DEF.sm ABL-sky take.down.PFV-3pmS-M ‘They took down the bananas from overhead (lit. sky).’ (249) yə-gʷrage berəsəβ akəna-m. ATTR-Gurage nationality ascend.PFV[.3smS]-M ‘He brought (lead) the Gurage people up [the hill].’ (250) asəna-xʷ-ɨm ə-tʒəppər-te. take.to.PFV-1sS-CV.M 1sS-return.IPFV-FUT.DEF ‘I will take it [there] and come back.’

101 (251) tʼəppʷə-m acənəw-i-m-tanə xɨ wəlando skin.PFV.IPS-CV.M bring.PFV.IPS-3smO-CV.M-LINK DEM welando kətəfʷ-i-m acənəw-i-m bar-ə-m. hash.PFV.IPS-3smO-CV.M bring.PFV.IPS-3smO-M say.PFV-3smS-M ‘He said: “They skinned and brought it, then they hashed welando¹⁰⁹ and brought it”.’

As mentioned above, some verbs of (transfer o) possession, even though not intransitive, can also form causatives with a-. The same is true with verbs of ingestion, which might be classified as a special case of verbs of (transfer o) possession. In both cases, there is a recipient who in a broad sense is the new possessor of an entity that is transferred; consider that both asəccʼə ‘give to drink’ and atʼəβətʼ ‘give to hold’ in (252) contain the verb ‘give’ in their English translations.

(252) Direct causatives of verbs of ingestion and transfer of possession bəna ‘eat’ → aβəna ‘feed’ səccʼə ‘drink’ → asəccʼə ‘give to drink’ tʼəppʷə ‘suck’ → atʼəppʷə ‘suckle’ cəffər ‘take a mouthful’ → acəffər ‘give a mouthful’ sɨyə ‘buy’ → asyə ‘sell’ tʼəβətʼ ‘grasp, hold’ → atʼəβətʼ ‘give to hold’ bətta ‘take’ → aβətta ‘give in marriage’

As for the marking of (primary) objects on the causatives of these non-intransitive verbs, there are cases where the theme and other cases where the recipient occurs as suffixes on the verb (i.e. only one at a time). In (253), it is the recipients -no ‘them’ and -naxə ‘you (sm)’ that occur as object suffix on the verb aβəna ‘feed’. In contrast, with asyə ‘sell’ in (254) it is the theme that is represented by the suffixed primary object -n ‘it’.

(253) a. cʼɨza t-aβəra-no-e ʃə-c-ɨm. drug 3sfS-feed.IPFV-3pmO-PURP want.PFV-3sfS-M ‘She wanted to feed them drugs (i.e. poison).’ b. zɨx asa atʼəməd-ə-m-ta y-aβəna-naxə mɨs this fish catch.PFV-3smS-CV.M-LINK REL-feed.PFV[.3smS]-2smO man y-axə yəməcʼərəʃa tʼəlat-axə-w. ATTR-2sm worstѦ enemy-2smPOSS-COP.3smS ‘This man, who cought and fed you fish, is your worst enemy.’ (254) a. yə-kəbbədə abba-ta bet-we asəyə-n ATTR-K. father-3smPOSS house-DEF sell.IMP[.2smS]-3smO bʷar-ə-n-ɨm? say.PFV-3smS-3smO-M ‘Did Kebbede’s father tell him to sell the house?’

¹⁰⁹A meal made of raw meat with spices and butter (Leslau 1979c: 652).

102 b. be, məngɨst-u asəyə-n yə-war-ə-n. no government-COP.3smS sell.IMP[.2smS]-3smO REL-say.PFV-3smS-3smO ‘No, the government told him to sell it.’ (Lit.: No, it is the government who told him “Sell it!”)

On the one hand, the occurrence of the object suffixes is conditioned by the definiteness of the referents (↗ 4.4.2). Thus, in (253) the themes cʼɨza ‘drug’ and asa ‘fish’ are (generic) indefinite and therefore unmarked, whereas the recipients are definite. In contrast, in (254) it is the theme betwe ‘the house’ that is definite and therefore marked on the verb (and the recipient is not specified at all). On the other hand, however, it also seems that different verbs allow or favor different alignments. In (255), for example, the theme zɨ dɨnica ‘these potatoes’ is definite and there is no recipient in form of an overt noun. Thus it could be thinkable that the 3sm object suffix refers to the potatoes. Nevertheless, the normal (if not the only possible) reading of this sentence is that “Kebbede fed these potatoes to him”. Conversely, example (254) above showed that a primary object suffix on asyə ‘sell’ refers to the theme, so that the recipient has to be expressed by means of a benefactive object as in (256).

(255) kəbbədə zɨ dɨnica awəna-n-ɨm. K. DEM potato feed.PFV[.3smS]-3smO-M ‘Kebbede fed him these potatoes.’ (*‘Kebbede fed these potatoes [to somebody].’) (256) kəbbədə dɨnica asyə-la-m. K. potato sell.PFV[.3smS]-BEN.3sf-M ‘Kebbede sold potatoes to her.’ (also ‘Kebbede sold potatoes for/in behalf of her’.)

3.7.3 Causativizer at- (indirect causativization) The causativizer at- typically forms causatives of transitive verbs. Nevertheless it is relatively free in use and can in principle occur with all types verbs including intransitive ones. As illustrated in (257), there are indirect causativizations from verbs of motion, perception, emotion or ingestion up to transitive verbs with highly agentive subjects like dənəg ‘beat’. Note that verbs with an initial vowel a – as the last three examples in (257) – can form causatives with at- only on formal grounds, regardless whether direct causativization with a- is expected or not.

103 (257) Examples of indirect causative derivations notʼ ‘run’ → atrotʼ ‘force to run’ ʒor ‘go round’ → atʒor ‘force to go round’ sənəf ‘be afraid’ → atsɨyənəf ‘scare someone’ ~atsenəf tənə ‘swear’ → attənə ‘have s.o. take an oath’ bəna ‘eat’ → atβəna ‘cause to eat, make eat’ xar ‘know’ → atxʸər ‘inform (“cause to know”)’ səppər ‘break’ → atsəppər ‘cause to break’ ʃəkət ‘fix, prepare’ → atʃəkət ‘have s.th. repaired’ ʒəppər ‘return’ → atʒəppər ‘get well, let s.o. return’ nəgəd ‘touch’ → atrəgəd ‘make touch’ kətəf ‘grind, hash’ → atkʸətəf ‘have s.th. ground, hashed’ dənəg ‘beat’ → addənəg ‘make beat, cause to beat’ aʃʃə ‘see’ → atyəʃʃə ‘show’ akək ‘scratch’ → atyəkək ‘itch (“make scratch”)’ ~atekək attər ‘spend the night’ → atyəttər ‘make spend the night; leave s.th. for next day’

Syntactically, at- introduces a new argument, the causer, which appears as the new subject of the causativized verb. In example (258b), this is the explicitly mentioned pronoun xʷɨt ‘he’. In turn, the old subject gɨyəxʷɨta ‘the dog’ in (258a) becomes the causee and appears as (additional) object of the causativized verb in (258b).

(258) a. gɨyə-xʷɨta fɨrank-əta bəna-m. dog-DEF.sm money-3smPOSS eat.PFV[.3smS]-3smO-M ‘The dog ate his money.’ b. xʷɨt yə-gyə-xʷɨta fɨrank-əta atwəna-n-ɨm. 3sm DAT-dog-DEF.sm money-3smPOSS make.eat.PFV[.3smS]-3smO-M ‘He made the dog eat his money.’

In causatives with at- there is a causer that makes someone else (the causee) do something. It is the causee that executes the action expressed by the basic verb and not the causer. Due to the fact that in contrast to causatives with a- there is this third actant “between” the causer and the action, derivations with at- are labeled indirect causativization. Other terms found in the literature are directive, distant, mediated, factitive or causee-controlled (Kulikov 2001: 892). It is not always necessary to explicitly mention the causee. In the derived sentence (259b) the old subject (xʷɨt ‘he’) is not present at all. Nevertheless it is understood that it is not the new subject (ɨyya ‘I’) that did the beating themself but that there is another person, the (implicit) causee, that did it.

104 (259) a. (xʷɨt) yə-mɨʃt-əta dənəg-ə-na-m. 3sm DAT-wife-3smPOSS hit.PFV-3smS-3sfO-M ‘He beat his wife.’ b. (ɨyya) yə-mɨʃt-əta addənəg-xɨ-na-m. 1s DAT-wife-3smPOSS make.beat.PFV-1sS-3sfO-M ‘I let beat his wife.’ / ‘I had his wife beaten.’

Ueno (2004) discusses at-causatives (in Chaha) and identifies two possibilities for the causee to appear, as accusative causee (260a) or as oblique causee with bə- (260b). Her examples are repeated here (transcription and glossing adapted).

(260) Chaha (Ueno 2004: 110) a. yə-raxel yə-mʷət-ə angaca atkʼəpən-nə-ya-m. DAT-R. REL-die.PFV-3smS cat make.bury.PFV-1pS-3sfO-M ‘We had Rachel bury the dead cat.’ b. yə-mʷət-ə angaca bə-raxel atkʼəpən-ne-m. REL-die.PFV-3smS cat INSTR-R. make.bury.PFV-1pS.3smO-M ‘We had the dead cat buried by Rachel.’

I do not have enough Gumer data to illustrate both possibilities (at least partly due to the fact that sentences with too many overt participants are not easily accepted), except example (261) with an oblique causee marked by bə-.

(261) (ɨyya) yə-wənəxʷə-na mɨʃt bə-mɨs-əxʸta 1s ATTR-neighbor-1sPOSS woman INSTR-man-3sfPOSS addənəg-xɨ-na-m. make.beat.PFV-1sS-3sfO-M ‘I had my neighbor’s wife beaten by her husband.’

As for the object suffixes on the causative forms, note that in (260b) the 3sm object suffix refers to the new object, the causee, whereas in (259b) it is the patient rather than the causee that is occurring on the verb as 3sf object suffix -na, i.e. the same referent as in the underived form in (259a). Also note that the use of object suffixes is largely conditioned by definiteness, animacy and other factors rather than the semantic role of the arguments as causee or patient (↗ 3.12.2). Concerning the semantics, there are several nuances that indirect causatives can express. The main meaning distinction is between (generally) making or letting a third intermediate person do an action – including the possibility to leave out the causee, i.e. having an action done (by someone) as in (259b) – and (more specifically) force someone to do something as in (258b). In this regard, consider sentence (262) as an example that can have both readings, i.e. letting or permit- ting to spend the night (for example as a guest) and forcing to spend the night (for example by the police).

(262) at kərə atyəttər-əβo-ndə-m. one day make.spend.the.night.PFV-3smS-1pO-M ‘They let/made us spend the night for one day.’

105 Further, an assistive or cooperative meaning (‘help to bring about’, ‘assist at bringing about’, cf. Kulikov 2001: 892) is attested in some verbs as for example atkʼattʼər in (263).¹¹⁰

(263) yunus yə-fədlu bet t-iy-arəʃʃ-i wədərə Y. ATTR-F. house TEMP-3S-build.house.IPS-3smO rope atkʼattʼər-ə-m. make.knot.PFV-3smS-M ‘When one was building Fedlu’s house, Yunus assisted knotting ropes.’

The causer is often human (or animate) acting more or less consciously as in the examples above, but it can just as well be an abstract concept, circumstances, etc. as it is exemplified in (264)-(266).

(264) yunus be bər-ot-əta atzakkʼʷər-ə-m. Y. no say-INF-3smPOSS make.chatter.PFV-3smS-1pO-M ‘Yunus saying no made [people] speak a lot.’ (265) yə-yunus ojə bɨzə səβ atmaccə-m. ATTR-Y. gossip many person make.angry[.3smS]-M ‘Yunus’s gossip made many people angry.’ (266) zənga-ta atsɨyənəf-ə-ndə-m. affair-3smPOSS scare.PFV-3smS-1pO-M ‘The situation scared us.’

As mentioned earlier, at- typically combines with transitive verbs. When it occurs with intransitive verbs, however, there is usually a contrast between direct causatives with a- and indirect causatives with at-. With verbs of (undirected) motion like notʼ ‘run’ it seems that the direct causative arotʼ ‘make run’ does not convey the meaning of forcing. Rather the focus lies on the movement as such which is facilitated or brought about by some cause. This is in a way comparable to causatives of verbs expressing states and change of states. Thus arotʼ as in (267) could be paraphrased as ‘causing that someone or something is in or enters the state of being in motion’.

(267) yə-xər-ə kʼar y-arotʼ-no. REL-become.PFV-3smS THING 3smS-make.run.IPFV-3pmO ‘Something makes them run.’

Hetzron (1977: 73) describes this meaning of not forcing as “make it possible / necessary for someone to do something” and mentions the verbs arəkʸəm ‘allow riding, make it possible to ride’ vs. atrəkʸəm ‘force to ride, order to ride’. Further, also take notice that the typical use of such direct causatives happens with ani-

¹¹⁰Another example is atrassa ‘help to lift’ derived from nəssa ‘lift’. Presumably the assistive meaning is only possible with inserted -a-. In the case of atk’attʼər ‘assist knotting’ this is not visible since the basic verb is kʼattʼər ‘knot’, a type C verb that already has an a.

106 mals as causee (268)-(270), expressing again that one puts the animals in the state of moving rather than telling (i.e. forcing) them to move.

(268) kʷɨtara y-arotʼ. chicken 3smS-make.run.IPFV ‘He chases chickens (he makes chickens run).’ (269) kʷɨtara y-aβərɨr. chicken 3smS-make.fly.IPFV ‘He makes chickens fly away (he flushes chickens).’ (270) fərəz y-aʒor. horse 3smS-make.go.round.IPFV ‘He makes horses go around in a circle.’

In contrast, indirect causatives with attend to be interpreted as ‘forcing someone to move’, or in other words the focus of the causer lies on the causee (in order that they move) rather than on the action of moving.

(271) aman y-atrotʼ-no. A. 3smS-make.run.IPFV-3pmO ‘Aman is chasing them.’

Consider also the comparable situation of the verb of motion ʒor ‘go round, turn’ (272) and its causative derivations. First, the direct causative in (273), featuring an inanimate causee (road), can be paraphrased as ‘cause that [the road] turns’ or ‘cause that [the road] is in the state of being turned’, and the indirect causative where the causer makes people perform the action of turning or going around (274).

(272) gəβya t-e-sər-o-m kənə-nyə ʒor-o-m. market TEMP-NEG.3S-arrive.IPFV-pmS-ALSO right-DIR turn.PFV-3smS-M ‘They turned to the right before they reached the market.’ (273) tə-x-m-e aʒor-xu-m-tanə ɨndəgəna b-atʼatʼ ema ABL-DEM-ALSO-GOAL make.turn.PFV-2pmS-CV.M-LINK againѦ LOC-A. road acənə-xu-m-tanə tə-drɨjjɨt ema dəməd-xu-m. bring.PFV-2pmS-CV.M-LINK COM-organization road join.PFV-2pmS-M (Speaking about the construction of a road:) ‘From there you turned [the road], and then you brought it to the road of Atat, then you connected [it] with the road of the organization.’ (274) əcʼɨr-xʷɨt atʒor-ə-m gəffər-ə-no-m. fence-DEF.sm make.go.round.PFV-3smS-CV.M release.PFV-3smS-3pmO-M ‘He forced (or made, or let) them to go around the fence and released them.’

With inchoative-stative verbs, the form with at- has to be considered the causative of the transitive direct causative with a- rather than a derivation “directly” from the intransitive basic verb. Thus, as demonstrated with bəssər ‘cook

107 (), become cooked, become ripe’ (275), the direct causative renders the verb transitive (276) and the indirect causative with at- expresses that someone makes another person perform the transitive event (277).

(275) ambər bəssər-ə-m. spinach cook.PFV-3smS-M ‘The spinach (got) cooked.’ (276) adot-əna ambər aβəssər-əc-ɨm. mother-1sPOSS spinach cook.PFV-3sfS-M ‘My mother cooked spinach.’ (277) ɨyya y-adot-əna ambər atβəssər-xɨ-na-m. 1s mother-1sPOSS spinach cook.PFV-1sS-3sfO-M ‘I had my mother cook spinach.’

3.7.3.1 Causation of reciprocity The infix -a- in combination with at-stems expresses causation of reciprocity (“cause to  each other”). A few examples are listed in (278) in comparison to other forms belonging to the same verb root.

(278) a. addənəg ‘cause to hit’ addanəg ‘cause to hit each other’ b. təwakka ‘fight each other’ atwakka ‘cause to fight each other’ c. nəkʸəm ‘ride’ atrakʸəm ‘pile up objects atop another (“cause to ride on each other”)’ d. atrəgəd ‘make someone touch someone’ atragəd ‘cause to touch one another’ e. atrəkəβ ‘hand over, cause (i.e. help) to find’ atrakəβ ‘introduce (“cause to find each other”)’ As Banksira (2000: 37) points out, causation of reciprocity requires a plural object.

(279) y-atraxɨβ-no 3smS-introduce.JUS-3pmO ‘let him introduce them’

108 3.8 Verbs without basic stem (prefix-necessitating or ‘bound’ stems) There are verbs that do not possess an underived basic stem. Petros (1994) discusses these prefix-necessitating verbs (in Chaha) and provides a presumably exhaustive list (Petros 1994: 1234f.).¹¹¹ There are verbs that occur with one (280) or two (281) of the prefixes tə-, a- and at- (there is no instance of a verb with a- and at- only).

(280) tə-*məccə — təməccə ‘be suitable’ *aməccə — *atməccə — a-*kʷəʃʃə — *təkʷəʃʃə — akʷəʃʃə ‘remove fibres from əssət’ *atkʷəʃʃə — at-*takkər — *tətakkər — *atakkər — attakkər ‘chase away’ (281) tə- & a- *kʼʸəppər — təkʼʸəppər ‘accept, receive’ akʼʸəppər ‘hand over, pass over’ *atkʼʸəppər¹¹² — tə- & at- *kʼaw — təkʼaw ‘drink coffee’ *akʼaw — atkʼaw ‘give coffee to drink’ Verbs that have only one ‘bound’ form can by and large be considered lexicalized items, but at least in some cases one can find a remnant meaning of intransi- tivity/reflexivity with tə- and transitivity/causativity with a- or at-. An exception are some verbs of ‘involuntary body movement’ (Petros 1994: 1221f.) which are intransitive despite the fact that they feature the prefix a-.¹¹³

¹¹¹Some of the prefix-necessitating verbs in Petros’s list in fact have a basic stem. However, in such cases the meaning of the verb with prefix should not be considered a valence-changing derivation from the basic stem. Rather these verbs possess their own (unpredictable but usually comprehensible) semantics (↗ 3.9). There are also some verb stems listed that are reciprocal/ frequentative forms for which there are basic underived stems, for example təsmamma ‘agree’ from səmma ‘hear’, but also without predictable semantic connection. ¹¹²From this verb root there is also a form with at- but with additional -a-, i.e. atkʼʸappər ‘respond, talk back, exchange words’. ¹¹³Some verbs of this group are experiencer verbs that realize their argument as primary object (↗ 3.12.2.1).

109 (282) axəna ‘shout’ (cf. xəyat ‘shouting’) agəssa ‘belch’ awzassa ‘perspire’ (cf. wɨzat ‘sweat’)

In the case of the prefix-necessitating verbs that occur in an opposition of tə- versus a- or at-, the forms are rather valency encoding than valency changing. This means that tə- expresses the intranstive and a- or at- the transitive meaning, whereby it is not decidable whether tə- reduces or a- increases the valency (remember that normally tə- reduces and a- increases the valency of the basic stem).

(283) təxəttər ‘dress onesel’ ↔ axəttər ‘dress someone’¹¹⁴ təxrəpəp ‘cover oneself ↔ axrəpəp ‘cover someone with cloth’ with cloth’ (284) təmar ‘learn’ ↔ atmər ‘teach’ təkʼaw ‘drink coffee’ ↔ atk’aw ‘give to drink coffee’ Some pairs of verbs as the two in (285) do not alter the (total) number of arguments. Instead, in the case of the transfer verb in (286) the role of the subject changes from goal (with tə-) to source (with a-). In the second verb pair with the meanings ‘follow’ and ‘precede’ (i.e. ‘make follow’) the subject and object change their role as follower and followed (287) (cf. Petros 1994: 1223f.).

(285) təkʼʸəppər ‘accept, receive’ ↔ akʼʸəppər ‘hand over, pass over’ təxʸəttər ‘follow’ ↔ axʸəttər ‘precede’ (286) a. aβdo xʷet kʷəncɨwə (tə-lula) təkʼʸəppər-ə-m. A. two small.pot ABL-L. receive.PFV-3smS-M ‘Abdo received two pots (from Lula).’ b. aβdo (yə-lula) xʷet kʷəncɨwə akʼʸəppər-ə-(na-)m. A. DAT-L. two small.pot pass.PFV-3smS-(3sfO-)M ‘Abdo passed two pots (to Lula).’ (287) a. gərəd ərc təxʸəttər-əc-ɨm. girl boy follow.PFV-3sfS-M ‘A girl follows a boy.’ b. gərəd ərc axʸəttər-əc-ɨm. girl boy precede.PFV-3sfS-M ‘A boy follows a girl (~a girl precedes a boy).’

¹¹⁴There is a basic stem of the same root √xdr, namely xəttər, but with a specialized meaning ‘thatch a roo’ (↗ 3.5.3).

110 3.9 Note on unpredictable and specialized meanings Some (derived) stems express a specialized meaning which is not predictable from the semantics of the basic stem (or the corresponding root) plus the derivational prefix. There are verb forms where the semantic connection is completely lost or at least obscured, even to the extent that they cannot easily be considered as belonging to the same verb root. Another instance are verbs that have gained additional meanings through semantic extension. Examples of verb roots where there is no transparent link between underived and derived stem are √frkʼkʼ with fɨrəkʼəkʼ ‘remove layers of plant one by one’ and təfrəkʼəkʼ ‘be arrogant, be careless’ or √gfr with gəffər ‘let go, release’ and agəffər ‘bring down’. In the case of the root √xdr there is a basic stem xəttər translating as ‘thatch a roo’ and two derived stems axəttər and təxəttər meaning ‘dress (/)’. There is little doubt that ‘thatch’ and ‘dress’ are etymologically connected sharing the common semantics ‘cover’. Nevertheless, axəttər and təxəttər ‘dress’ are not derivations of xəttər ‘thatch’. Instead, the latter nowadays has a specialized separate meaning which is not part of a three-way valency changing set Ø vs. tə- vs. a-. Another example is axʷə ‘leak’ which strictly speaking is not the causative of xʷə ‘spill, pour’ (but there is atxʷə ‘make spill’). Two instances of verbs that have gained specialized meanings by semantic extension (i.e. they still coexist with the transparent meaning) are agəppa ‘put in, bring in’ which also means ‘marry’ (said of a man who “lets enter” the wife into his home) (288) or awəttʼa ‘take out, bring out’ which can – among other things – also mean ‘name, appoint’ (289).

(288) bə-tkɨnət agβ-ot wəxe kʼar ã-xər-ə. LOC-childhood marry-INF good THING NEG-become.PFV-3smS ‘Getting married in childhood is not good.’ (289) zax səβ ɨnnɨm-əxʷna […] at-at-at-at-at yə-ʒɨr danə DEM person all-3pmPOSS [...] one-one-one-one-one ATTR-zhir judge awəttʼ-o-m. name.PFV-3pmS-M ‘These people all […] named one zhir-judge each (i.e. five in total).’

Further, note that there are (derived) stems containing the reciprocal infix -a- that do not express reciprocity, or at least this meaning is not obvious anymore. The verb atwana in (290), for example, relates to the verbs awəna ‘put’ and cona ‘sit’ and could be rephrased as ‘cause to sit on each other’. Nevertheless, the extended meaning ‘make a plan’ can hardly be understood as a reciprocal action (cf. Leslau 1979c: 654). Also the quadriradical verbs in (291) do not represent a meaning of reciprocity. Moreover, they do not possess a basic stem and cannot be compared to forms without -a-.

(290) √UrA atwana ‘pile up in a special way; make a plan’

111 (291) √frtʼr təfrattʼər ‘jump around, bounce’ √rfrf tərfanəf ‘become dirty’ arfanəf ‘make dirty’

Finally, there are also verbs with the shape of a Frequentative (↗ 3.10) that do not straightforwardly express the repeated action of the verb they are derived from.

(292) √kʼmr kʼəmmər ‘delouse’ kʼɨmammər ‘do lousy, do not whole-heartedly’ √kʼβr kʼəppər ‘bury, plant’ akʼβappər ‘cover fire with ashes to keep it alive’ √dβr dəppər ‘add’ tədβappər ‘be wrong’

3.10 Frequentatives Reduplication of the medial radical is commonly known in Ethiosemitic languages as the ‘Frequentative’. It is a derivation that conveys the notion of intensity or repetition added to the basic meaning of the verb. The formation is fairly productive with (mostly transitive) triradical verbs. Quadriradicals and verbs already containing a reduplicated radical, i.e. with the shape 122 (↗ 3.5.2.1) or 113 (↗ 3.5.2.3), cannot form Frequentatives (cf. Rose 2007: 410). Frequentative derivations are possible with basic as well as all derived stems (t(ə)-, a-, at-). An overview of the bases of the basic sound triradicals and derived stems is given in tables 48 and 49. Note that not many verbs could be collected, especially for the derived stems, and that there is some puzzling variation in the Jussive concerning gemination and palatalization, a problem which remains to be resolved by more reliable data.¹¹⁵

Type    A  12ə2ə3 ~ 12a2ə3 12ə23 ~ 12a23 12ə23 ~ 12a23 12ə23~ 12a23 B  12ə2ə3 ~ 12a2ə3 +  12ə23 ~ 12a23 +  12ə23 ~ 12a23 +  12ə23~ 12a23 +  C  12a2ə3 12a23 12a23

Table 48: Templates of Frequentatives of sound triradical verbs

¹¹⁵The general problem of variation of gemination vs. non-gemination (but also palatalization vs. depalatalization) in the Jussive is also mentioned in Hetzron (1977: 70). He states that “[t]here seems to be interdialectal variation”. Probably one could even go further and say that there can be idiolectal variation, fostered by the multitude of dialects some speakers are exposed to.

112 Prefix Type    t(ə)- A  t(ə)-12a2ə3 t-12a2ə3 t(ə)-12a2ə3 / (t(ə)-12a2ə3) B  t(ə)-12a2ə3 + t-12a2ə3 + t(ə)-12a2ə3 – / (+) a- A  a-12a2ə3 a-12a23 a-12a23 / (a-12a23) B  a-12a2ə3 + a-12a23 + a-12a23 + at- A  at-12a2ə3 at-12a23 at-12a23 B  at-12a2ə3 + at-12a23 + at-12a23 +

Table 49: Templates of Frequentatives of derived stems of sound triradical verbs

3.10.1 Frequentatives of basic stems The Frequentative is derived from triradical verbs by reduplication of the penultimate radical and insertion of either the vowel ə or a between the two reduplicated radicals (1ə2ə3 → 12ə2ə3/12a2ə3). Table 50 shows verbs of types A, B and C and their corresponding Frequentative derivations in the Perfective. Note that verbs of type B that feature the ‘palatalized’ vowel e (like metʼəs ‘detach’) completely lack it in the Frequentative.

Type Root     A √sβr səppərə sɨβəppərə / sɨβappərə ‘break’ B √dβr jəppərə jɨβəppərə / jɨβappərə ‘finish’ √mgr məkkʸərə mɨkʸəkkʸərə / mɨkʸakkʸərə ‘set fire’ √mtʼs metʼəsə mɨtʼətʼəsə / mɨtʼatʼəsə ‘detach’ C √zβt zapətə zɨβapətə ‘get lost’

Table 50: Examples of Frequentative derivations ()

Generally, all Frequentatives of basic stems can occur with inserted ə as well as a (table 51). The two variants do not seem to bear any semantic differences and are freely interchangeable. However, the forms with a are much more common and the few verbs of type C – if forming Frequentatives at all – appear to avoid the variant with ə.¹¹⁶ The derived Frequentatives are quadriracals with largely the same patterns as the underived quadriradical types E and F. In particular, it is only the penultimate radical that is subject to mutation, whereas the first half of the reduplicated pair occurs in its unchanged form. Thus, the Frequentative reveals the underlying nature of the penultimate radical especially of the regular type B verbs, which have mutation in all bases. Consider for instance the g of the Frequentative jɨgəkəm ‘hit

¹¹⁶No Frequentatives of the small group of type D verbs could be found. It is not clear if they are impossible due to fromal restrictions or if they are theoretically possible but simply not in use.

113 with fist repeatedly’, which is derived from jəkəm ‘hit with fist’ (√dgm), in contrast to the k of the Frequentative cɨkəkkər ‘cook repeatedly, cook several things’, which is derived from cəkkər ‘cook’ (√tkr). In the Jussive, the Frequentative differs slightly from regular quadriradicals in three aspects. First, the vowels ə or a are always placed between the reduplicated radicals, i.e. between the second and the third radical, whereas in quadriradicals of type E the place of the vowel ə is after the first radical. Second, Frequentatives of type B verbs do not depalatalize in the Jussive.¹¹⁷ Third, the bases of the Jussive often show mutation/gemination, but as mentioned above there is some considerable variation.¹¹⁸ Apart from that, note that the three verb types converged to basically the same templates.

Type Root    A  √sβr sβəppər sβə(p)pr sβə(p)pr ‘break ~ sβa(p)pər ~ sβa(p)pr ~ sβa(p)pr repeatedly’ B  √dgm jgəkəm jgəkm jgəkm ‘hit with fist ~ jgakəm ~ jgakm ~ jgakm repeatedly’ C  √zβt zβapət zβapt zβapt ‘get lost several times’

Table 51: Examples of Frequentatives bases

As with all reduplicated verb forms palatalization and labialization usually affect both doubled consonants, as in the Imperative 2sf (293a) and the Impersonal (293b).

(293) a. jɨgəkɨm ‘bash repeatedly (sm)ǃ’ jɨgʸəkʸɨm ‘bash repeatedly (s)ǃ’ b. sɨβəppər-ə-m ‘he broke repeatedly’ sɨwəppʷər-i-m ‘one broke repeatedly’ It has been mentioned above that Frequentatives express the notion of intensity and most commonly repetition. Repetition does not mean that an action is done ‘again’ but that it happens more than once. This can come about in basically two different ways: either the action itself is carried out several times or it is carried out on several objects. Naturally, the third logical option is a combination of the first two, i.e. the action is carried out several times on several object. Thus, in particular with unspecific or indefinite objects, there are usually three possible readings (294).

¹¹⁷Though, for the type B verb məkkʸər ‘set fire’ with palatalized second radical I have recorded the depalatalized Frequentative Jussive yəmkəkkɨr ‘set fire repeatedly’ (cf. Perfective mɨkʸəkkʸərəm and Imperfective yɨmkʸəkkʸɨr). ¹¹⁸For example, I have recorded Jussive bases ʒɨβaβr and ʒβa(p)pr of the verb ʒɨβappər ‘turn over’ (derived from ʒəppər √zβr ‘return’).

114 (294) wədərə mɨtʼatʼəs-ə-m. rope detach.FREQ.PFV-3smS-M ‘He detached the/a rope several times.’ ‘He detached (the/several) ropes.’ ‘He detached (the/several) ropes several times.’

When the direct object is clearly singular as it is the case with the 3sm definite article -xʷɨt in (295) or, in absence of an overt noun phrase, for example also the 3smO suffix -n in (296) and (297), the only conceivable reading is that the action is carried out several times on the same object.

(295) yə-tʼay-xʷɨt tʼɨβaβətʼ-ə-m aʃʃə-n-ɨm. DAT-sheep-DEF.sm grasp.FREQ.PFV-3smS-CV.M see.PFV[.3smS]-3smO-M ‘He checked [the quality o] the sheep by touching/grasping it several times.’ (296) kʼɨmaməs-ə-m cə-n-ɨm. taste.FREQ.PFV-3smS-CV.M leave.PFV[.3smS]-3smO-M ‘He tasted it a few times and left it.’ (297) əram-xʷɨt mɨsəsəx-ə-m jəppʷər-ə-n-ɨm. cow-DEF.sm ruminate.FREQ.PFV-3smS-CV.M finish.PFV-3smS-3smO-M ‘The cow finished it ruminating several times.

Since Gumer does not feature nominal plural marking, nouns are ambiguous in terms of their number (more precisely nouns unless marked with a definite article, ↗ 4.2). In such cases plurality can be discerned through the Frequentative form of the verb. In (298) mena ‘work’ must refer to more than one job due to the Frequentative yəjβaprɨm (as opposed to the underived form yədəprɨm) ‘he should ... and finish’.

(298) mena-ta yə-jβapr-ɨm tɨmɨrtɨβet yə-wər. work-3smPOSS 3smS-finish.FREQ.JUS-CV.M school 3smS-go.JUS ‘He should go to the school to finish his jobs.’

The following pair of examples illustrates the common correlation between plural object and Frequentative (299a) on the one hand and singular object and underived verb form (299b) on the other hand.

(299) a. gɨβɨr-xɨno ʃɨkakət-ə-m awəna-no-m. stuff-DEF.pm organize.FREQ.PFV-3smS-CV.M put.PFV[.3smS]-3pmO-M ‘He keeps his stuff (several objects) well organized.’ (More literally: ‘He has put the stuff having organized them.’) b. gɨβɨr-xʷɨt ʃəkət-ə-m awəna-n-ɨm. stuff-DEF.sm organize.PFV-3smS-CV.M put.PFV[.3smS]-3pmO-M ‘He keeps his stuff (one object) well organized.’ (More literally: ‘He has put the stuff having organized it.’)

115 In the second sentence (299b), the object gɨβɨrxʷɨt ‘the stuf’ could possibly not only be understood as consisting of a single object but also as ‘stuff as a whole’ in the sense that it actually does consist of several discernible objects but is perceived or treated as one entity. Anyhow, a comparable situation arises in (300) with one of the very few nouns that have a (suppletive) plural form like əram ‘cow’ and əray ‘cows’ (see table 82). The latter expressing plurality, it expectedly occurs with the Frequentative of ʃəkkər ‘change’ as in (300a) with the interpretation that the action is carried out on each individual member of the group of əray ‘cows’. However, it can also stand with the underived verb form as in (300b). Consequently the action is carried out only once which in turn means that əray, though consisting of a number of individual cows, is treated as one entity.

(300) a. əray ʃɨgakkɨrǃ cows change.FREQ.IMP[.2smS] ‘Change the cowsǃ’ (for example the place of grazing of each cow) b. əray səgɨrǃ cows change.IMP[.2smS] ‘Exchange the cowsǃ’ (for example with sheep)

Finally, in some cases the Frequentative is better understood as lexicalized form with distinct (though related) semantics rather than a derivation of the basic verb. The verb ʒɨβappər in (301), for example, is formally a Frequentative of ʒəppər ‘return ()’ but it simply translates as ‘turn over, turn upside down’. The important point here is that it does not mean that the action of turning over wussa happened several times (but note that a frequentative reading ‘turn over several times’ is also possible).

(301) yunus wɨssa ʒɨβappər-ə-m. Y. wussa turn.over.PFV-3smS-CV.M ‘Yunus turned over the wussa-bread.’

3.10.2 Frequentatives of derived stems As already mentioned above, also stems derived with t(ə)-, a- or at- can form Fre- quentatives. Like the basic triradical verb, they double the medial radical and the vowel a (or occasionally also ə) is inserted between the reduplicated consonants (for example tə-1ə2ə3 → tə-12a2ə3). Tables 52-54 show example verbs of type A and B of each stem (no verbs of type C and D could be found). Again, note that all Frequentative derivations tend to converge to one common pattern of mutation/gemination in all bases. The Jussive bases often showing variation, there are also forms without mutation attested as in the example of type A in table 53. Fur- ther, the Jussive of type B verbs usually depalatalizes in t-stems, but not in the other two.

¹¹⁹Sometimes also the first of the doubled consonants appears in the mutated form as in təmkakkər √mxr ‘advise each other’. This is also reported by Leslau (1950) cited in Hetzron (1977: 71).

116 Type Root    A √mxr t-mkakkər t-mkakkər t-mkakkər ‘advise each other’¹¹⁹ B √kʼβr t-kʼʸβappər t-kʼʸβappər t-kʼβappər ‘receive from each other’

Table 52: Examples of Frequentatives bases of derived verbs with t(ə)-

Type Root    A √kʼβr a-kʼβappər a-kʼβappr a-kʼβaβr ‘cover fire with ashes’ B √dgr a-jgakkər a-jgakkr a-jgakkr ‘be troublesome’

Table 53: Examples of Frequentatives bases of derived verbs with a-

Type Root    A √rgd at-rgagəd at-rgagd at-rgagd ‘make many people get in touch with each other’ B √kʼrβ at-kʼʸranəβ at-kʼʸranβ at-kʼʸranβ ‘serve several things’

Table 54: Examples of Frequentatives bases of derived verbs with at-

Occasionally the vowel between the doubled medial radical is ə instead of a. It seems that this is the preferred case with weak verbs.

(302) √rsI arəʃʃə ‘plait the hair’ → tərʃəʃʃə ‘plait each other’s hair’ Verbs with a weak second radical reduplicate the first radical to form the Frequen- tative. The only examples available, all of them reciprocals with tə-, are listed in (303). The first two verbs belong to type A-1A3-Ø, which shows palatalization in some forms, as they also do in the derived stems (note the vowel e in təβeβər).

(303) √xAr təxʸər ‘be known’→ təxʸəxʸər ‘get to know one another’ √βAr təβer ‘be said’ → təβeβər ‘say to each other’ √kʼIf ɨnkʼʸəf ‘embrace’ → tənkʼʸəkʼʸəf ‘embrace each other’ ~tənkʼʸakʼʸəf Verbs that have a weak first radical A also copy the second radical and feature the vowel ə or a between the doubled pair, with this type of verb seemingly having a strong preference for ə. As for the initial vowel of the root, it does not surface as a anymore in the Frequentative (304). Banksira (2000: 66) segments this verb as t-əgəgəz, i.e. the slot of the first radical is not empty but filled with ə as the remnant of the initial a.

(304) √Agz agəz ‘help’ → təgagəz ‘help each other’ ~təgəgəz

117 Further, in the Imperfective and Jussive bases of the derived Frequentative of such verbs like aʃʃə √AzI ‘see’ (305) the initial radical A leaves a trace insofar as the prefix t- is geminated followed by the epenthetic vowel ɨ in the place of the first radical.

(305)  təʒəʒə¹²⁰ ‘look at each other’  yɨttɨʒəʒ (*yɨdʒəʒ)  yəttɨʒəʒ (*yɨdʒəʒ)

Frequentatives of t-stems derived from basic stems are used to express reciprocal actions carried out more than once, typically by several people simultaneously and/or everybody acting on each other as in (306) (the basic stems being ɨnkʼʸəf ‘embrace’ and bar ‘say’).

(306) arβət kabort yə-txəttər-əma gɨred tənkʼʸəkʼʸəf-əma-m səlam four coatѦ REL-dress.PFV-3pfS girls embrace.each.other.PFV-3pfS-CV.M greeting təβeβər-əma-m. say.to.each.other.PFV-3pfS-M ‘Four girls wearing coats hugged and greeted each other.’

A reciprocal state of affairs is also treated as Frequentative when more than one pair of twos act on each other parallelly as in (307)-(308).

(307) yə-xʷet-xʷet gʷɨnər yɨ-tɨrsəs-əma. DAT-two-two hair 3S-plait.each.other.IPFV-pfS ‘They plait their hair each other in pairs.’ (308) yə-xʷet-xʷet-əxnəma kʼɨmar yɨ-tkʼɨmammər-əma. DAT-two-two-3pfPOSS louse 3S-delouse.each.other.IPFV-pfS ‘They delouse each other in pairs.’

With verbs that do not have a basic stem like təkʸəppər ‘receive, accept’, the Fre- quentative does not necessarily express reciprocity. Rather, the relation between təkʼʸəppər in (309) and təkʼʸβappər in (310) is the same as between a (“normal”) basic stem and its Frequentative, though in this case one could translate təkʼʸβappər with ‘receive from each other’.

(309) kubbayya tɨ-t-iβ-na təkʸəppər-əc-na-m. cupѦ TEMP-3sfS-give.IPFV-3sfO receive.PFV-3sfS-3sfO-M ‘When she gave her a cup, she received (accepted) it from her.’ (310) xʷet dengʸa arβət gɨred cəkʷəlat təkʼʸβappər-o-m tʼəβətʼ-o-m. two boys four girls chocolate receive.FREQ.PFV-3pmS-M hold.PFV-3pmS-M ‘Two boys [and] four girls exchanged chocolate (and kept it).’

¹²⁰The expected form would be təʒəʃʃə with geminated and mutated penultimate radical, but I have only recorded təʒəʒə. I do not have an explanation for that.

118 Yet another case represents təʒβappər ‘turn around ()’ in (311). Rather than being the Frequentative reciprocal of ʒəppər or təʒəppər ‘return (/)’, this is the anticausative of ʒɨβappər ‘turn around, turn over ()’ which has already been described as lexicalized form with its own specialized (though related) semantics (see example (301)).

(311) təʒβappər-ə-m y-aʒ-ɨn. turn.around.PFV-3smS-CV.M 3smS-see.IPFV-3smO ‘Turning around, he sees him.’

Just as basic transitive verbs like the ones in (298)-(300) above do, also derived stems form Frequentatives when there are multiple objects. Compare the singular object in (312) with the Frequentative in example (313) containing two objects.

(312) yɨ-sətʼ-əβo kʼar an-atkʼʸənəβ-əc banə. 3S-drink.IPFV-pmS THING NEG-bring.near.PFV-3sfS AUX.PT ‘She had not served (lit. brought near) anything to drink.’ (313) murida ʃay-ɨm kʼawa-m atkʼʸɨranəβ-əc-ɨm. M. tea-ALSO coffee-ALSO bring.near.PFV-3sfS-M ‘Murida served (lit. brought near) tea and coffee.’

Frequentatives of a-stems express repeated actions the expected way as in (314) where the verb awəttʼa ‘take out’ has several objects.

(314) ɨmɨr-ɨm əcʼə-m awtʼattʼa-m. stone-ALSO wood-ALSO take.out.FREQ.PFV[.3smS]-M ‘He took out stones and wood.’

It seems, however, that Frequentatives of a-stems tend to adopt specialized semantics quite often. Thus, awtʼattʼa can also mean ‘investigate’, or the Frequenta- tive akʼβappər, a derivation from kʼəppər ‘bury’, is used most of the time together with ɨsat ‘fire’ in the very specific meaning ‘cover fire with ashes to keep it alive’ (rather than the literal translation ‘bury repeatedly’).

(315) kʼəppər ‘bury’ akʼβappər ‘cover fire with ashes to keep it alive (“bury repeatedly”)’ (316) ɨsat akʼβaβɨrǃ fire bury.FREQ.JUS[.2smS] ‘Cover the fire with ashesǃ’

Consider also the following few Frequentatives of a-stems (in comparison to non- Frequentatives) expressing a meaning that is semi-specialized in the sense that it is not predictable but still retraceable.

(317) ajəggər ‘trouble ()’ ajgakkər ‘be troublesome (“make trouble repeatedly”)’

119 (318) ʃətta ‘smell ()’ aʃətta ‘smell ()’ aʃtatta ‘snif’ (319) ʒəppər ‘retrun ()’ aʒβappər ‘take/carry back and forth (“make return repeatedly”)’

Finally, note the case of the Frequentativ aftʼattʼər that coexists with afəttʼər, but with basically the same meaning ‘be fast’ (apart from the other meaning ‘prepare food’). If there is a semantic difference it seems to be very subtle.

(320) afəttʼər ‘boil (coffee), prepare food; hurry, be fast’ aftʼattʼər ‘be fast, do something quickly’ (321) afəttʼər-xə-m cənə-xə-m? hurry.PFV-2smS-M come.PFV-2smS-M ‘Did you come quickly (soon)?’ (322) aftʼattʼər-xʷ-ɨm cot-xʷ-ɨm ə-ʒʒəppər-te. do.fast.PFV-1sS-CV.M make.PFV-1sS-CV.M 1sS-return.IPFV-FUT.DEF ‘I will do it quickly and come back.’

3.11 Subject marking Verbs are obligatory marked with subject affixes added to the verb bases indicating person, number, and gender. The Perfective receives suffixes only, whereas the Imperfective and Jussive have prefixes and additional suffixes. Table 55 summarizes the subject markers of all three TAM forms. The position of the base is indicated by three dots.

   1s …-xʷ ə-… n-… 2sm …-xə t-… … 2sf …-xʸ t-…+ …+ 3sm …-ə y-… yə-… 3sf …-əc t-… t-… 1p …-nə n-…-nə n-…-nə 2pm …-xu t-…-o …-o 2pf …-xma t-…-əma …-əma 3pm …-o y-…-o yə-…-o 3pf …-əma y-…-əma yə-…-əma  …++ y-…++ (yə-)…++

Table 55: Subject markers

120 All TAM forms feature the common three persons plus a “fourth person” called Impersonal (), a hallmark of Gurage languages. The Impersonal is special in terms of form and use and is treated in section 3.11.5. The other persons all occur as singular and plural. The second and third persons distinguish between masculine and feminine gender. The second person feminine singular forms all contain palatalization, either as -xʸ in the subject marker of the Perfective or as palatalization within the Imperfective or Jussive base (↗ 3.11.4). Note that the consonantal prefixes and suffixes beginning with a consonant often occur with an epenthetic vowel ɨ according to the epenthesis rules (↗ 2.3.2). Compare, for example, the subject markers pre- or suffixed to a vowel of the verb base with the markers affixed to a consonant as in (323).

(323) a. bəna-xma-m vs. cot-xɨma-m eat.PFV-2pfS-M work.PFV-2pfS-M ‘you ate’ ‘you worked’

b. t-aʒ vs. tɨ-ʒor 2smS-see.IPFV 2smS-go.around.IPFV ‘you see’ ‘you go around’

Some of the subject markers have allomorphs in combination with object markers, subordinating prefixes, negation or with weak verbs (↗ 3.11.6).

3.11.1 Perfective conjugation The suffixes of the Perfective are different for each person. However, all second persons contain the element x, and both plural feminine forms end with ma. Table 56 shows the complete Perfective conjugation of a regular type A verb. Take notice that these forms here are bare forms that do not occur in isolation like this, but only in combination with further affixes, i.e. the main verb marker -m (↗ 3.18.1.1), subordinators (↗ 4.7), and/or negation an- (↗ 3.15). For the Impersonal see section 3.11.5.

  1 kəfət-xʷ kəfət-nə 2m kəfət-xə kəfət-xu 2f kəfət-xʸ kəfət-xɨma 3m kəfət-ə kəfət-o 3f kəfət-əc kəfət-əma  kəfʷəc(-i)

Table 56:  of √kft ‘open’

121 3.11.2 Imperfective conjugation The Imperfective has prefixes and additional suffixes in the plural. Third person is marked by y-, whereas t- stands for second person, but typical of Semitic languages t- is also used in the third person singular feminine. The first persons do not share the same prefix, ə- standing for singular and n- for plural (but see section 3.11.6.3 for the allomorph n- of ə-). As for the suffixes in the plural, there is -nə in the first person, echoing the nasal of the prefix, whereas in the second and third persons, -o marks masculine and -əma feminine. Finally, note that the Imper- sonal has a default third person prefix y- (↗ 3.11.5). Table 57 shows the complete Imperfective conjugation of a regular type A verb.

  1 ə-kəft nɨ-kəft-ɨnə 2m tɨ-kəft tɨ-kəft-o 2f tɨ-kəfc tɨ-kəft-əma 3m yɨ-kəft yɨ-kəft-o 3f tɨ-kəft yɨ-kəft-əma  yɨ-kəfʷc(-i)

Table 57:  of √kft ‘open’

3.11.3 Jussive conjugation The subject markers of the Jussive in table 58 resemble the ones of the Imperfec- tive, but there are the following differences: the first person singular is marked by n-, analogous to the first person plural; the third persons (except for the singular feminine t-) have the prefix yə- with an additional ə; and finally the second persons, which are the Imperatives, do not have a prefix at all. In the Impersonal there are two possibilities: a regular Jussive with a default subject prefix yə- and an Im- perative without prefix. According to Banksira (2000: 254) these two variants are in free variation.

  1 nɨ-kɨft nɨ-kɨft-ɨnə 2m kɨft kɨft-o 2f kɨfc kɨft-əma 3m yə-kɨft yə-kɨft-o 3f tɨ-kɨft yə-kɨft-əma  yɨ-kɨfʷc(-i) ~kɨfʷc(-i)

Table 58:  of √kft ‘open’

122 When the Jussive base is used to build other forms than the ‘bare’ Jussive, the subject affixes are the same as in the Imperfective shown in table 59. This is the case with the Indefinite Future (↗ 3.18.5), in Negation (↗ 3.15.3) and in the apodosis of irreal conditional clauses (↗ 4.7.3.5). Hetzron (1996: 102) calls these forms “archaic jussive” (i.e. the affixes of the actual Jussive in table 58 are innovations).

  1 ə-kɨft nɨ-kɨft-ɨnə 2m tɨ-kɨft tɨ-kɨft-o 2f tɨ-kɨfc tɨ-kɨft-əma 3m yɨ-kɨft yɨ-kɨft-o 3f tɨ-kɨft yɨ-kɨft-əma  yɨ-kɨfʷc(-i)

Table 59: ‘Archaic’  of √kft ‘open’

3.11.4 Formation of feminine singular Verbs denoting second person feminine singular subjects contain a palatal element, i.e. there is a palatalized consonant or a fronted vowel. This palatalization goes back to the suffix *-i, but in Gumer there is no separable morpheme anymore.¹²¹ In the Perfective, the suffix for 2sf is -xʸ, as opposed to 2sm -xə.

(324) nəkəβ-xʸ-ɨm vs. nəkəβ-xə-m find.PFV-2sfS-M find.PFV-2smS-M ‘you (s) found’ ‘you (sm) found’

In the Imperfective and Jussive we find the palatalization of a consonant in the base or, if not possible, the fronting of a vowel. As trigger of this palatalization one can assume a high vocoid -I ‘suffixed’ to the verb base (cf., for example, Banksira 2000: 191). Examples (325)-(349) each show the 2sf Imperfective and Imperative, and in brackets the corresponding masculine forms for comparison. In bases ending with a velar (k, kʼ, g, x) (325) or alveolar obstruent (t, tʼ, d, s, z) (326) it is always this final consonant that is palatalized, irrespective of the properties of the preceding radicals.

(325) tɨ-sərkʼʸ (tɨ-sərkʼ) ‘you steal’ sɨrkʼʸ (sɨrkʼ) ‘stealǃ’ tɨ-ʒangʸ (tɨ-ʒang) ‘you leave’ ʒangʸ (ʒang)¹²² ‘leave!’

¹²¹See Lowenstamm (2000) for a (formal) analysis of the Chaha 2nd feminine singular formation.

123 tɨ-fʷəxʸ (tɨ-fʷəx) ‘you wipe’ fʷɨxʸ (fʷɨx) ‘wipeǃ’ (326) tɨ-gʸənʒ (tɨ-gʸənz) ‘you cut in a big slice’ gʸɨnʒ (gʸɨnz) ‘cut in a big sliceǃ’ tɨ-wərj (tɨ-wərd) ‘you descend’ wɨrəj (wɨrəd) ‘descendǃ’ tɨ-rocʼ (tɨ-rotʼ) ‘you run’ nocʼ (notʼ) ‘run!’ tɨ-akʸəʃ (t-akʸəs) ‘you joke’ akʸəʃ (akʸəs) ‘joke!’ tɨ-kəfc (tɨ-kəft) ‘you open’ kɨfc (kɨft) ‘open!’

The palatalized form of r is y, realized as i after consonants, i.e. Cɨy → Ci (cf. (58)).

(327) tɨ-səβi (tɨ-səβɨr) ‘you break ()’ sɨβi (sɨβɨr) ‘breakǃ ()’ tɨ-ʒəppi (tɨ-ʒəppɨr) ‘you return ()’ zəppi (zəppɨr) ‘return ()’ tɨ-tʼəkʼi (tɨ-tʼəkʼɨr) ‘you hide ()’ tʼɨkʼi (tʼɨkʼɨr) ‘hide! ()’ tɨ-xʷrəxʷi (tɨ-xʷrəxʷɨr) ‘you take out the earwax’ xʷərxʷi (xʷərxʷɨr) ‘take out the earwaxǃ’

If the palatalized r→y follows the vowel ə, this combination is realized as e, i.e. Cəy → Ce (cf. example (53)). This is for example the case in the Jussive of verb type A2 (328), which has the template 12ə3, or in the Imperfective and Jussive of derived stems with prefixed t(ə)- or at- (329), or in the Jussive of verbs with a missing penultimate radical like bar ‘say’ and wər ‘go’ (330).

(328) tɨ-rəβi (tɨ-rəβɨr) ‘you live’ nɨβe (nɨβər) ‘live!’ t-adi (tɨ-adɨr) ‘you spend the night’ əde (ədər) ‘spend the nightǃ’ (329) tɨ-tgʸəkkʸe (tɨ-tgʸəkkʸər) ‘you look up’ təgəkke (təgəkkər) ‘look upǃ’ t-atme (t-atmər) ‘you teach’ atme (atmər) ‘teachǃ’

¹²²For the Jussive of ʒanəg ‘leave, go away’ also ʒarg and zarg are attested (but not *zang).

124 (330) tɨ-βi (tɨ-βɨr) ‘you say’ be (bər) ‘sayǃ’ t-ay (t-ar) ‘you go’ we (wər) ‘go’

If the palatalized r→y follows the vowel o, we have the combination oy.

(331) tɨ-ʒoy (tɨ-ʒor) ‘you go around’ zoy (zor) ‘go around!’

Further, a final l (which can occur in loans) is also palatalized to y/i.

(332) tɨ-ci (tɨ-cɨl) ‘you can’ cay (cal) ‘can!’

In case the base final consonant is already palatalized (i.e. in verbs with a final radical I or U, which both cause palatalization of preceding radicals, ↗ 3.4.2.1.3), the feminine and masculine forms do not differ (333).

(333) tɨ-rəkʸ (tɨ-rəkʸ) ‘you throw’ ərəgʸ (ərəgʸ) ‘throw!’ tɨ-txʸakʼʸ (tɨ-txʸakʼʸ) ‘you boast’ təxʸakʼʸ (təxʸakʼʸ) ‘boast!’ tɨ-βəxʸ (tɨ-βəxʸ) ‘you cry’ bɨxʸ (bɨxʸ) ‘cry!’ tɨ-mac (tɨ-mac) ‘you get angry’ mac (mac) ‘get angry!’ tɨ-arəʃ (tɨ-arəʃ ) ‘you weave, you plait’ arʃ (arʃ ) ‘weaveǃ, plaitǃ’

If the base final consonant is not palatalizable, i.e. if it is a labial (p, b, β, f, m) or the nasal n, palatalization floats further left. Either, the vowel i is inserted between the second but last and the final consonant of the base (or replaces the epenthetic ɨ) (334), or when there is already the vowel ə or a between these two consonants, these vowels are raised to e or ɛ, respectively (335).

(334) t-ariβ (t-arɨβ) ‘you milk’ əriβ (ərɨβ) ‘milk!’ tɨ-kətif (tɨ-kətf ) ‘you hash’ kɨtif (kɨtf ) ‘hash!’

tɨ-gʸətim (tɨ-gʸətɨm) ‘you lend’ gətim (gətɨm) ‘lend!’

125 tɨ-wəssin (tɨ-wəssɨn) ‘you decide’ (< ) wəssin (wəssɨn) ‘decide!’ (335) tɨ-tkʼʸəneβ (tɨ-tkʼʸənəβ) ‘you are (become) near’ təkʼəneβ (təkʼənəβ) ‘be(come) nearǃ’ tɨ-ʃeβ (tɨ-ʃəβ) ‘you pull’ ʃɛβ (~ ʃeβ)(ʃaβ) ‘pull!’ tɨ-cʼen (tɨ-cʼən) ‘you give birth; you father’ tʼen (tʼən) ‘give birth!; father!’

However, when the second but last consonant is a velar, there is no insertion or raising of a vowel, but this velar is palatalized (336). Furthermore, a velar can even be palatalized in third but last position in case the second but last consonant is r (337), but these forms alternate with insertion of i in the manner of the verbs in (334) above.

(336) tɨ-jəkʸɨm (tɨ-jəkɨm) ‘you bash’ dəkʸɨm (dəkɨm) ‘bashǃ’ tɨ-rəxʸɨβ (tɨ-rəxɨβ) ‘you find’ nɨxʸəβ (nɨxəβ) ‘find!’ (337) tɨ-kʼʸərm~tɨ-kʼərim (tɨ-kʼərm) ‘you insult’ kʼʸɨrm~kʼɨrim (kʼɨrm) ‘insult!’

Velars that are already labialized usually become palatalized, too, losing hereby their labialization. However, at least in the case of nəkkʼʷə ‘roar’, vowel raising and retention of labialization is also attested.

(338) tɨ-fkʼʸən (tɨ-fɨkʼʷən) ‘you whistle’ fɨkʼʸən (fɨkʼʷən) ‘whistle!’ tɨ-kʷrəkʸɨm (tɨ-kʷrəkʷɨm) ‘you give a blow with the knuckles’ kʷərkʸɨm (kʷərkʷɨm) ‘give a blow with the knucklesǃ’ tɨ-xʸə (tɨ-xʷə) ‘you spill’¹²³ tɨ-rəkʼʸ ~tɨ-rekʼʷ (tɨ-rəkʼʷ) ‘you roar’ nɨkʼʸ ~nikʼʷ (nɨkʼʷ) ‘roar!’

In verbs with a final radical A, palatalization does not directly target the base final a but floats further left to the preceding consonants treating them the same way as the regular base final consonants described above; the base final a, however, is changed to ə in all cases. Thus, rightmost velars (k, kʼ, g, x) (339), alveolar obstruents (t, tʼ, d, s, z) (340), and the liquid r (341) are palatalized.

¹²³The Imperative is xʷay for both masculine and feminine.

126 (339) t-afəkʸə (t-afəka) ‘you remove’ afkʸə (afka) ‘removeǃ’ tɨ-wəgʸə (tɨ-wəga) ‘you stab, fight’ wɨgʸə (wɨga) ‘stabǃ, fightǃ’ (340) tɨ-βəjə (t-βəda) ‘you take away’ bɨjə (bɨda) ‘take awayǃ’ tɨ-rəʃə (tɨ-rəsa) ‘you lift’ nɨʃə (nɨsa) ‘liftǃ’ (341) tɨ-zəyə (t-zəra) ‘you sow’ zɨyə (zɨra) ‘sow!’

If the consonant preceding the base final a is not palatalizable, i.e. if it is a labial (p, b, β, f, m) or the nasal n, palatalization floats further left. As with the regular verbs, only velars are palatalized in this position (342), whereas other consonants remain unaffected (even if they are palatalizable elsewhere). Instead, the vowel i is inserted (or replaces the epenthetic ɨ) between the two consonants or the vowel ə is raised to e (343).

(342) tɨ-gʸəfə (tɨ-gəfa) ‘you push’ gʸɨfə (gɨfa) ‘push!’ tɨ-kʼʸəβə (tɨ-kʼəβa) ‘you oil’ kʼʸɨβə (kʼɨβa) ‘oilǃ’ tɨ-kʼʸrəppə (tɨ-kʼrəppa) ‘you snap of’ kʼʸəmbə (kʼəmba) ‘snap off!’ (343) tɨ-tʼeβə (tɨ-tʼəβa) ‘you take away’ tʼiβə (tʼɨβa) ‘skinǃ’ tɨ-semə (tɨ-səma) ‘you listen’ simə (sɨma) ‘listen!’

In reduplicated verbs, palatalization affects both the base final and the doubled consonant of the pair. Example (344) shows verbs with final and (345) with total reduplication. Note that with doubled r only the base final r is palatalized (346).

(344) t-akʸɨkʸ (t-akɨk) ‘you scratch’ əkʸɨkʸ (əkɨk) ‘scratch!’ tɨ-gəjj¹²⁴ (tɨ-gədd) ‘you pierce’ gɨjj (gɨdd) ‘pierce!’ t-aʃɨʃ (t-asɨs) ‘you sweep’ əʃɨʃ (əsɨs) ‘sweep!’

¹²⁴Or with epenthetic ɨ between the identical final radicals (see example (84)): tɨ-gəjɨj (tɨ-gədɨd) ‘you

127 (345) tɨ-kʃəkʃ (tɨ-ksəks) ‘you dash to bits’ kəʃkɨʃ (kəskɨs) ‘dash to bitsǃ’ tɨ-sxʸəsxʸ (tɨ-sxəsx)¹²⁵ ‘you grind slightly with pestle’ (346) tɨ-βəri¹²⁶ (tɨ-βərɨr) ‘you fly’ bɨre (bɨrər) ‘flyǃ’ tɨ-ʃraʃʃi (tɨ-ʃraʃʃɨr) ‘you level the ground’ ʃɨraʃi (ʃɨraʃɨr) ‘level the ground!’

If the final doubled consonants are not palatalizable, the vowel i is inserted between them, but palatalization of a preceding velar is also possible (347). Likewise also totally reduplicated verbs receive an inserted i when the base final consonant is not palatalizable (348).

(347) tɨ-kʼəfif (tɨ-kʼəfɨf ) ‘you cut edges’ kʼɨfif (kʼɨfɨf ) ‘cut edges!’ tɨ-kʼʸəbb (tɨ-kʼəbb) ‘you shave’ kʼʸɨbb (kʼɨbb) ‘shave!’ (348) tɨ-tβətiβ (tɨ-tβətɨβ) ‘you tie up’ tətiβ (tətɨβ) ‘tie upǃ’

Also in Frequentatives (↗ 3.10) palatalization always affects both parts of the doubled medial radical (349).

(349) tɨ-jgʸəkʸɨm (tɨ-jgəkɨm) ‘you bash repeatedly’ jɨgʸəkʸɨm (jɨgəkɨm) ‘bash repeatedlyǃ’

3.11.5 Formation of Impersonal The Impersonal is formed by labialization and palatalization within the bases rather than by an overt and separable affix. This formation goes back to a suffix *-u originally denoting plural masculine (cf. Hetzron 1977: 83). Synchronically, one can assume a vocoid -U ‘suffixed’ to the verb base, whose features [round] and [high] trigger labialization and palatalization in the preceding base, respectively (cf. Banksira 2000: 206ff.). These sound changes are the same in all of the three basic TAM forms. The following examples show the Impersonals of the Perfective, Imperfective and Jussive, and in brackets the 3sm forms for comparison. Note that the -i in the Impersonal is the 3sm, which is present by default (↗ 3.11.5); for the final main verb marker -m in the Perfective see section 3.18.1.1.

pierce’, gɨjɨj (gɨdɨd) ‘pierceǃ’ ¹²⁵The Imperative is səsxʸ (səsx) with deletion of a radical (see table 26 and examples (151)-(152)). ¹²⁶Other forms I have recorded do not involve palatalization of the final r but insertion or suffixation of i: tɨ-βərir and tɨ-βərri.

128 The feature [round] of -U floats left and targets the first labializable consonant. Thus, base final labials (350) and velars (351) are all labialized. Note that velars are also palatalizable, but in the Impersonal they are labialized only.

(350) nəkəw-i-m (nəkəβ-ə-m) ‘find’ yɨ-rəxw-i (yɨ-rəxɨβ) yə-nxəw-i (yə-nxəβ) kətəfʷ-i-m (kətəf-ə-m) ‘hash, chop’ yɨ-kətfʷ-i (yɨ-kətf ) yə-kɨtfʷ-i (yə-kɨtf ) gʸətəmʷ-i-m (gʸətəm-ə-m) ‘lend’ yɨ-gʸətmʷ-i (yɨ-gʸətɨm) yə-gətmʷ-i (yə-gətɨm) (351) adəgʷ-i-m (adəg-ə-m) ‘throw down’ y-adgʷ-i (y-adɨg) y-ədgʷ-i (y-ədɨg) sənəkʼʷ-i-m (sənəkʼ-ə-m) ‘steal’ yɨ-sərkʼʷ-i (yɨ-sərkʼ) yə-sɨrkʼʷ-i (yə-sɨrkʼ) manəxʷ-i-m (manəx-ə-m) ‘capture’ yɨ-manxʷ-i (yɨ-manx) yə-manxʷ-i (yə-manx)

If the base final consonant is not a labial or velar, labialization floats further left until it finds a suitable host, as the second but last consonant in (352), or even further left as in (353).

(352) kəfʷəc-i-m (kəfət-ə-m) ‘open’ yɨ-kəfʷc-i (yɨ-kəft) yə-kɨfʷc-i (yə-kɨft) (353) kʼʷənəʃ-i-m (kʼənəs-ə-m) ‘begin’ yɨ-kʼʷərʃ-i (yɨ-kʼərs) yə-kʼʷɨrʃ-i (yə-kʼɨrs)

Base final coronal obstruents (t, tʼ, d, s, z) are palatalized due to the feature [high] of -U. This is illustrated in (352)-(353) and in (354). The latter examples also show that labialization is totally absent if there is no labializable consonant in the base at all.

(354) wacʼ-i-m (watʼ-ə-m) ‘swallow’ yɨ-wəcʼ-i (yɨ-wətʼ) yə-wacʼ-i (yə-watʼ)

129 oj-i-m (od-ə-m) ‘tell’ y-uj-i (y-ud) y-oj-i (y-od) agʸagʸəʒ-i-m (agʸagʸəz-ə-m) ‘be proud, show of’ y-agʸagʸʒ-i (y-agʸagʸz) y-agʸagʸʒ-i (y-agʸagʸz)

Other than the palatalization trigger -I of the 2nd person singular feminine (↗ 3.11.4), the -U of the Impersonal does not palatalize r (355).

(355) ʃəkkʷər-i-m (ʃəkkər-ə-m) ‘change’ yɨ-ʃəkʷr-i (yɨ-ʃəkkɨr) yə-səgʷr-i (yə-səgɨr)

Furthermore, palatalization triggered by -U does not float, i.e. only the base final consonants are palatalized. The base final n in the Jussive in example (356) is not palatalizable, but also the preceding t, even though a palatalizable consonant, remains unchanged.

(356) y-atən-i (y-atən) ‘let him bring (let one bring)’

The base final a in verbs with a final radical A does not block palatalization. The palatalizable consonants preceding a are treated like base final consonants, as for example the t→c in (357). The base final a, however, is changed to ə (as in the palatalization process of the 2nd person singular feminine, ↗ 3.11.4). When followed by the default 3sm suffix -i (which is by far the most frequent occurrence of an Impersonal), ə and i fuse to the vowel e.

(357) /nəccə-i-m/ → nəcce-m (nətta-m) ‘separate (layers of əssət)’ /yɨ-rəcə-i/ → yɨ-rəce (yɨ-rəta) /yə-ncə-i/ → yə-nce (yə-nta)

Take note that the base final a always changes to ə, also when no palatalization of a consonant takes place (358).

(358) /səmmʷə-i-m/ → səmmʷe-m (səmma-m) ‘hear’ /yɨ-səmʷə-i/ → yɨ-səmʷe (yɨ-səma) /yə-smʷə-i/ → yə-smʷe (yə-sma) /bʷənə-i-m/ → bʷəne-m (bəna-m) ‘eat’ /yɨ-wərə-i/ → yɨ-wəre (yɨ-βəra) /yə-wrə-i/ → yə-wre (yə-βra)

All verbs with base final ə have an alternative way of forming the Imper- sonal Perfective base which is used very frequently in Gumer. They feature an epenthetic or additional β in the place of the missing final consonantal radical which then becomes regularly labialized to w, for example bəkkʸə ‘cry’ and fʷəʃʃə

130 ‘fart’ (359). In the Imperfective and Jussive of these two verb, which have no base final ə, there is (normally) no additional β→w. In contrast, a verb like sɨyə √srAI ‘sell’ with base final ə in Imperfective and Jussive again features β→w (360).

(359)  bəkkə+β-i-m → bəkkəw-i-m (~ ¿bəkkʷ-i-m) ‘one cried’ fʷəssə+β-i-m → fʷəssəw-i-m (~ ¿fʷəʃʃ-i-m)¹²⁷ ‘one farted’  yɨ-βəxʷ-i (*yɨ-βəxʸəw-i, ‘one cries’ *yɨ-βəxəw-i) yɨ-fʷəʃ-i (*yɨ-fʷəʃəw-i, ‘one farts’ *yɨ-fʷəsəw-i)  yə-βxʷ-i (*yə-βxəw-i) ‘let one cry’ yɨ-fʷʃ-i (*yɨ-fʷʃəw-i) ‘let one fart’

(360)  yɨ-srə+β-i → yɨ-srəw-i¹²⁸ ‘one buys’  yə-sərə+β-i → yə-sərəw-i ‘let one buy’

In reduplicated verbs, labialization and palatalization always¹²⁹ occur on both consonants of the doubled pair, as in the following totally reduplicated verbs: (361) with final labialization, (362) with final palatalization, and (363) with non-final labialization and final palatalization. Example (364) shows the totally reduplicated verb sassa ‘be thin’ with a final A (√sAsA).

(361) tɨwətəw-i-m (tɨβətəβ-ə-m)¹³⁰ ‘tie up’ yɨ-twətw-i (yɨ-tβətɨβ) yə-tətw-i (yə-tətɨβ) (362) nɨʒənəʒ-i-m (nɨzənəz-ə-m) ‘nag’ yɨ-nʒənʒ-i (yɨ-nzənɨz) yə-nəʒnɨʒ-i (yə-nəznɨz) (363) kʷɨʃəkʷəʃ-i-m (kɨsəkəs-ə-m) ‘dash to bits’ yɨ-kʷʃəkʷʃ-i (yɨ-ksəks) yə-kʷəʃkʷɨʃ-i (yə-kəskɨs) (364) ʃaʃʃe-m (sassa-m) ‘be(come) thin’ yɨ-ʃaʃʃe (yɨ-sassa) yə-ʃaʃʃe (yə-sassa)

¹²⁷At times even fʷəʃʃəwim seems acceptable. ¹²⁸Or alternatively also yɨsre. ¹²⁹Occasional exceptions to the rule are possible, i.e. at times only the second consonant of the doubled pair is affected. ¹³⁰Also attested with gemination tɨwəttəw-i-m (tɨβəttəβ-ə-m).

131 The following two examples show verbs with final reduplication: (365) with labialization and (366) with palatalization.

(365) səkʷəkʷ-i-m (səkək-ə-m) ‘drive a peg’ yɨ-səkʷɨkʷ-i (yɨ-səkɨk) yə-skʷɨkʷ-i (yə-skɨk) (366) fʷəʒəʒ-i-m (fəzəz-ə-m) ‘be(come) better’ yɨ-fʷəʒʒ-i (yɨ-fəzz) yə-fʷʒəʒ-i (yə-fzəz)

Also in Frequentatives, i.e. in verbs with medial reduplication, labialization (367) and palatalization (368) affect both consonants.

(367) təʒwappʷər-i-m (təʒβappər-ə-m) ‘turn upside down ()’ yɨ-ʒʒɨβappər-i (yɨ-ʒʒɨwappʷər) yə-ʒʒɨwawər-i (yə-ʒʒɨβaβər) (368) aʃcacce-m¹³¹ (aʃtatta-m) ‘snif’ y-aʃcacce (y-aʃtatta) y-aʃcace (y-aʃtata)

Consonants that are already palatalized cannot be labialized in the Impersonal. The verb gʸəkkʸər ‘straighten out, arrange’ in (369), for example, has palatalized velars in the Perfective and Imperfective, which remain unchanged, whereas in the Jussive they are depalatalized (↗ 3.16) and therefore become open to labialization.

(369) gʸəkkʸər-i-m (gʸəkkʸər-ə-m) ‘straighten out, arrange’ yɨ-gʸəkʸr-i (yɨ-gʸəkkʸɨr) yə-gʷəkʷr-i (yə-gəkkɨr)

Velars that are already labialized do not change. However, note that the verb nəkkʼʷə ‘roar’ is probably the only such example (370), for which also palatalization is attested in the feminine singular (↗ 3.11.4).

(370) nəkkʼʷ-i-m (nəkkʼʷə-m) ‘roar’ yɨ-rəkʼʷ-i (yɨ-rəkʼʷ) yə-rkʼʷ-i (yə-rkʼʷ)

¹³¹Here, I have also recorded instances where only the last t is palatalized, i.e. aʃtaccem etc.

132 3.11.6 Allomorphs of subject markers 3.11.6.1 Verbs with weak final radical As illustrated in table 60, some of the bases of verbs with a weak final radical (i.e. without consonant as final radical) end in a vowel. If the final radical is A, the Perfective, Imperfective and Jussive bases end in a. If the final radical is I or U, the Perfective base ends in ə, whereas the Imperfective and Jussive bases do not feature a final vowel.

   A √gβA gəppa gəβa gβa ‘enter’ I √sxI səkkʸə səxʸ sxʸ ‘flee’ U √fkʼrU fkʼʷənə fkʼʷən fkʼʷən ‘whistle’

Table 60: Example bases of verbs with weak final radical

These base final vowels and suffixed subject markers beginning with a vowel interact with each other, which can bring about allomorphs in some cases. In the Perfective, the vowel of the subject suffixes 3sm -ə and 3sf -əc is absorbed by the base final vowel a (371) or ə (372), resulting in the allomorphs -Ø and -c.¹³²,¹³³

(371) 3sm *bəna-ə-m → bəna-Ø-m → bəna-m ‘he ate’ eat.PFV-3smS-M eat.PFV-3smS-M eat.PFV[.3smS]-M 3sf *bəna-əc-ɨm → bəna-c-ɨm ‘she ate’ eat.PFV-3sfS-M eat.PFV-3sfS-M (372) 3sm *cənə-ə-m → cənə-Ø-m → cənə-m ‘he came’ come.PFV-3smS-M come.PFV-3smS-M come.PFV[.3smS]-M 3sf *cənə-əc-ɨm → cənə-c-ɨm ‘she came’ come.PFV-3sfS-M eat.PFV-3sfS-M

Note that if the base final vowel is not a (371) but ə (372), one could also argue that it is the base final vowel that is absorbed (*cənə-ə-m → cən-ə-m). However, in analogy to the instances with a, where clearly the subject marker is deleted, it is more conclusive to assume that this is also the case in (372).

¹³²The zero morpheme -Ø is not represented in the examples elsewhere; in the glosses square brackets are used to indicate the ‘missing’ but implicitly known morpheme. ¹³³Banksira (1999b: 29, 2000: 243) assumes that the perfective bases of sound verbs also end in -ə (for example √kft → kəfətə-), thus the 3sm and 3sf subject suffixes (always) being -Ø and -c (for example kəfətə-Ø-m and kəfətə-c-ɨm). However, in all other cases (be it a subject suffix beginning with a consonant or a vowel) the verb base does not show a final ə (for example 2sm kəfət-xə-m), a fact which he “explains” by stating that ə is truncated when followed by CV or a vocoid. Banksira (1999b, 2000: 241ff.) needs all these assumptions for his analysis of the system of all Chaha subject affixes. In my view this is an over-analysis that poses problems rather than simplifying matters.

133 In contrast, the plural subject suffixes 3pm -o and 3pf -əma override the base final vowel, no matter if it is a (373) or ə (374).

(373) 3pm *səna-o-m → sən-o-m ‘they (m) arrived’ arrive.PFV-3pmS-M arrive.PFV-3pmS-M 3pf *səna-əma-m → sən-əma-m ‘they () arrived’ arrive.PFV-3pfS-M arrive.PFV-3pfS-M (374) 3pm *kʼʸəmmə-o-m → kʼʸəmm-o-m ‘they (m) won’ win.PFV-3pmS-M win.PFV-3pmS-M 3pf *kʼʸəmmə-əma-m → kʼʸəmm-əma-m ‘they () won’ win.PFV-3pfS-M win.PFV-3pfS-M

Similar to the singular forms above, one might argue that in the case of 3pf -əma after ə it is the vowel of the suffix that is dropped (*kʼʸəmmə-əma-m → kʼʸəmmə- ma-m, accordingly with an allomorph -ma). Nevertheless, in analogy to all other instances in the plural such an analysis would be rather inconsistent. It goes without saying that the same rule also applies for second and third persons plural of the Imperfective (375) and Jussive (376) bases that end in a.

(375) 2pm *tɨ-cona-o → tɨ-con-o ‘you (pm) sit’ 2S-sit.IPFV-pmS 2S-sit.IPFV-pmS 2pf *tɨ-cona-əma → tɨ-con-əma ‘you (p) sit’ 2S-sit.IPFV-pfS 2S-sit.IPFV-pfS 3pm *yɨ-cona-o → yɨ-con-o ‘they (m) sit’ 3S-sit.IPFV-pmS 3S-sit.IPFV-pmS 3pf *yɨ-cona-əma → yɨ-con-əma ‘they () sit’ 3S-sit.IPFV-pfS 3S-sit.IPFV-pfS (376) 2pm *tora-o → tor-o ‘sit downǃ (pm)’ sit.IMP-pmS sit.IMP-pmS 2pf *tora-əma → tor-əma ‘sit downǃ (p)’ sit.IMP-pfS sit.IMP-pfS 3pm *yə-tora-o → yə-tor-o ‘let them (m) sit down’ 3S-sit.JUS-pmS 3S-sit.JUS-pmS 3pf *yə-tora-əma → yə-tor-əma ‘let them () sit down’ 3S-sit.JUS-pfS 3S-sit.JUS-pfS

In addition to the ‘normal’ forms as seen in (373)-(376), there is a second variant that often occurs after bases with a final vowel: -əβo and -əβəma (instead of - o and -əma). Thus, as illustrated with some examples in (377), there are often two possibilities. Like their shorter forms, the longer allomorphs -əβo and -əβəma delete the final vowels of the verb base. Again, this is particularly evident with bases that end in a.¹³⁴

¹³⁴I have recorded only one instance where the base final a was not deleted: ellaβom ‘they coveted’.

134 (377) 3pm *wəttʼa-o-m → wəttʼ-əβo-m ~wəttʼ-o-m ‘they (m) ascend.PFV-3pmS-M ascend.PFV-3pmS-M went up’ 3pf *wəttʼa-əma-m → wəttʼ-əβəma-m~ wəttʼ-əma-m ‘they () ascend.PFV-3pfS-M ascend.PFV-3pfS-M went up’ 2pm *tɨ-cona-o → tɨ-con-əβo ~tɨ-con-o ‘you (pm) sit’ 2S-sit.IPFV-pmS 2S-sit.IPFV-pmS 2pf *tɨ-cona-əma → tɨ-con-əβəma~ tɨ-con-əma ‘you (p) sit’ 2S-sit.IPFV-pfS 2S-sit.IPFV-pfS 3pm *assə-o-m → ass-əβo-m ~ ass-o-m ‘they (m) saw’ see.PFV-3pmS-M see.PFV-3pmS-M 3pf *assə-əma-m → ass-əβəma-m~ ass-əma-m ‘they () saw’ see.PFV-3pfS-M see.PFV-3pfS-M

Very generally speaking, the variants with the additional əβ are in free distribution with the shorter ‘normal’ forms (i.e. with verb bases ending in a vowel). The choice seems to depend at least partially on idiolectal preferences, but some geo- graphical distribution might also play a role. Supposedly in villages closer to the Chaha area the longer forms are less frequent, but this claim has to be verified. Then again, however, there is a rather clear difference in frequency depending on the quality of the final vowel. As illustrated with wəttʼa and cona in (377), bases ending in a can occur with the longer allomorphs, but here the shorter forms are significantly more common, i.e. for instance səmma → səmm-o-m ‘they heard’ is clearly preferred over səmm-əβo-m. In contrast, verbs with bases ending in ə tend to occur with -əβo and -əβəma much more often, i.e. for example fɨkʼʷənə → fɨkʼʷən-əβo-m ‘they whistled’ is as at least as common as fɨkʼʷən-o-m if not the preferred form, up to the extent that some speakers would reject the shorter one. Above, the additional əβ is analyzed as part of the subject marker since it deletes the base final vowels a as do the shorter ‘normal’ forms. However, based on the (preferred) formation of the Impersonal (Perfective) of verbs with a base final ə (i.e. where the longer subject suffixes with β are rather frequent), one might also conclude that the β belongs to the base itself. The Impersonal does not have an overt subject suffix, but with said verbs there is in most cases a base final glide w, very much the same as if the verb had a final radical β which is labialized in the Impersonal (↗ 3.11.5). Compare the Perfective Impersonals of nəkəβ and bəkkʸə.

(378) √rxβ ‘find’ √βxI ‘cry’ 3sm nəkəβ-ə-m bəkkʸə-m 3pm nəkəβ-o-m bəkk-o-m~ bəkk-əβo-m  nəkəw-i-m bəkkəw-i-m

It is not clear if this was a performance mistake or a further (though less frequent) possiblity. (It was not intended to mean ‘he coveted to their detriment’, which would have the same form.)

135 On grounds of the Impersonal bəkkəwim and not knowing any other forms of the paradigm one would have to conclude that its base is bəkkəβ, and consequently the (longer form o) 3pm should then be analyzed as bəkkəβ-o-m (with the regular subject suffix -o). However there is also the equally possible shorter form bəkk-o-m based on bəkk(ʸə) (with depalatalization, ↗ 3.16), thus β is not an essential part of the base in the plural forms. Due to that, but above all due to the fact already exemplified above that -əβo/-əβəma delete the base final a of verbs like cona ‘sit’ (cona → con-o-m ~ con-əβo-m) these longer forms are treated here as allomorphs of the ‘regular’ subject markers -o/-əma. This analysis is supported by the fact that the forms with β occasionally even occur with verbs that have a consonantal final radical, for example wər-əβo-m instead of wər-o-m ‘they went’. This phenomenon being rather seldom in the Perfectives it occurs slightly more often in the Imperfective and Jussive, for example y-adr-əβo instead of y-adr-o ‘they spend the night’ or yə-tot-əβo instead of yə-tot-o ‘let them (pm) work’. The data seem to point out that these variations are only found with weak verbs, i.e. with verbs having a weak (i.e. non-consonantal) radical also in other positions than the last one. Sound verbs featuring only sound consonantal radical apparently cannot occur at all with the longer subject suffixes. Note that -əβo/-əβəma can be restricted in use or at least become the dispreferred choice when these forms coincide with a Perfective verb with 3sm -ə +  -β-. Compare for example the two different segmentations of cənəβom in (379).

(379) a. cən-əβo-m vs. b. cənə-βo-m come.PFV-3pmS-M come.PFV[.3smS]-MAL.3pm-M ‘they (m) came’ ‘he came to their (m) detriment’

The plural in (a) above is a perfectly usual form, but it could at times come in con- flict with the 3sm +  interpretation (b). In such cases speakers might tend to avoid ambiguities by opting for the shorter form cən-o-m. However, such misun- derstandings presumably do not occur very often, except where the context is not clear or missing. As for the weak verbs that do not have a base final ə such as wər ‘go’ (380), the use of -əβo/-əβəma is possible (a), but the malefactive reading (b) seems more natural in such a way that plural forms with -əβo/-əβəma generally are dispreferred. Thus, in contrast to verbs with a base final vowel (as for instance the above mentioned cənəβom ‘they came’), here the shorter subject markers are more common (i.e. wərom/wərəmam ‘they went’).

(380) a. ¿wər-əβo-m vs. b. wər-ə-βo-m go.PFV-3pmS-M go.PFV-3smS-MAL.3pm-M ‘they (m) went’ ‘he went to their (m) detriment’

If the plural forms -əβo/-əβəma do not completely coincide with 3sm + , they are again more acceptable and frequent since there is no ambiguity anymore. For instance, many verbs with base final vowels feature also palatalization which is blocked (or depalatalized) in the third persons plural of the Perfective (↗ 3.16).

136 Consider the example verb aʃʃə ‘see’ in (381) with palatalized ʃʃ vs. non-palatalized ss. Here, the form 3sm + malefactive aʃʃəβom (c) cannot be mistaken for 3pm, which is assəβom (b) (in contrast to cənəβom in (379) above).

(381) a. aʃʃə-m vs. b. ass-əβo-m~ass-o-m see.PFV[.3smS]-M see.PFV-3pmS-M ‘he saw’ ‘they (m) saw’ c. aʃʃə-βo-m see.PFV[.3smS]-MAL.3pm-M ‘he saw to their (m) detriment’ (*‘they saw’)

According to Banksira (2000: 233f., 243), the masculine plural marker -o goes back to *-əβU, i.e. following the regular sound processes the marker is derived as /-əβU/ → /-əβʷ/ → /əw/ → [o]. Thus, if this analysis is correct, the bilabial element β is not only an additional element of the allomorph -əβo of the masculine plural marker, but is underlyingly already part of the regular -o. Given that -əβo pre- dominantly occurs with verbs that have a base final vowel, one could then further assume that this β occupies or docks onto the empty slot of the missing consonantal radical. Consequently, it becomes to some extent also part of the base, which might serve as an explanation that in the Impersonal of such verbs it is virtually always present (i.e. in its labialized form w). The thesis that the masculine plural marker -o derives from *-əβU is supported by the fact that the plural feminine forms -əβəma are slightly less frequent than the masculine -əβo. For instance, cən- əβo-m tends to be interpreted as more natural than cən-əβəma-m ‘they came’, or the other way round the shorter form cən-əma-m seems somewhat more acceptable and consequently used more often than the corresponding short masculine cən-o-m. Nevertheless, even if the diachronic analysis *-əβU → -o as posited by Banksira is adequate, it does not account for the actual synchronic occurrence of - əβo. First of all, as seen above, there is also the feminine form -əβəma, furthermore these allomorphs with β are used in the Imperfective and Jussive, where there is no base final vowel ə (see table 60), and they even occur with verbs that do not have a weak final radical at all. Moreover, if -o is the direct result of *-əβU, why should the last vowel still be o in the longer form -əβo, and not, for example, u (*-əβu)? All in all it does not seem to be farfetched to hypothesize that -əβo origins in *-əβU (or similar) and occurred first as 3pm with (Perfective) verbs that have a base final vowel. Later, -əβ- must have been reanalyzed as an element to be used generally after base final vowels, thus yielding feminine -əβəma. Then, presumably, the use of both -əβo and -əβəma as plural markers of verbs with weak final radical expanded to the Imperfective and Jussive to the extent that nowadays it can even occur with all other weak verbs. This supposed development also reflects by and large the frequency of occurrence of the allomorphs with β.

137 3.11.6.2 Verb bases with initial vowel In the Imperfective, the 1s subject prefix ə- is not overt with verb bases beginning in a vowel. This is the case when the verb has a weak first radical (except U) (382), and with the causative prefixes a- and at- (383).

(382) *ə-agd → Ø-agd → agd ‘I tie’ 1sS-tie.IPFV 1sS-tie.IPFV [1sS.]tie.IPFV *ə-ella → Ø-ella → ella ‘I covet’ 1sS-covet.IPFV 1sS-covet.IPFV [1sS.]covet.IPFV *ə-ud → Ø-ud → ud ‘I tell’ 1sS-tell.IPFV 1sS-tell.IPFV [1sS.]tell.IPFV (383) *ə-acən → Ø-acən → acən ‘I bring’ 1sS-bring.IPFV 1sS-bring.IPFV [1sS.]bring.IPFV *ə-atmər → Ø-atmər → atmər ‘I teach’ 1sS-teach.IPFV 1sS-teach.IPFV [1sS.]teach.IPFV

The same is true for the ə of the third person marker yə- in the Jussive which is deleted when followed by a base beginning with a vowel. Again this is with verbs that do not have a consonantal first radical (384) – but note that if it is A, the Jussive bases (except Imperative) do not have an initial vowel – or with the causative prefixes a- and at- (385).

(384) *yə-od → y-od ‘let him tell’ 3sS-tell.JUS 3sS-tell.JUS *yə-ella → y-ella ‘let him covet’ 3sS-covet.JUS 3sS-covet.JUS (385) *yə-amʷəkʼ → y-amʷəkʼ ‘let him heat’ 3sS-heat.JUS 3sS-heat.JUS *yə-atəʒ → y-atəʒ ‘let him show’ 3sS-show.JUS 3sS-show.JUS

This does not only result in the allomorphy yə-~y- in the Jussive, but in some cases there is also a formal identity between Imperfective and Jussive, namely if the two bases happen to have the same shape.

(386) y-akʷəʃ vs. y-akʷəʃ (<*yə-akʷəʃ ) 3sS-remove.fibres.IPFV 3sS-remove.fibres.JUS ‘he removes fibres’ ‘let him remove fibres’ (387) y-atmər-əma vs. y-atmər-əma (<*yə-atmər-əma) 3S-teach.IPFV-pfS 3S-teach.JUS-pfS ‘they () teach’ ‘let them () teach’

138 3.11.6.3 1s Imperfective with prefixes In the Imperfective, the 1s prefix ə- changes to -n- when it is preceded by another prefix, i.e. the negation marker a- (388) on the one hand, and the subordinators t- (389) and b- (392) on the other hand. It hereby becomes formally the same as the 1s marker of the Jussive and the 1p marker of the Imperfective and Jussive, which are n- also without preceding prefix (see table 55).

(388) ə-xʸɨr → a-n-xʸɨr 1sS-know.IPFV NEG-1sS-know.IPFV ‘I know’ ‘I do not know’ (389) ə-cot → tɨ-n-cot 1sS-work.IPFV TEMP-1sS-work.IPFV ‘I work’ ‘when I work’

In verbs with an initial vowel, where the subject marker ə- is deleted, it appears again as -n- after the same prefixes.

(390) ar → a-n-ar [1sS.]go.IPFV NEG-1sS-go.IPFV ‘I go’ ‘I do not go’ (391) aʒ → tɨ-n-aʒ [1sS.]see.IPFV TEMP-1sS-see.IPFV ‘I see’ ‘when I see’ (392) ame → bɨ-n-ame [1sS.]do.IPFV TEMP-1sS-do.IPFV ‘I do’ ‘when I do’

The 1s -n- assimilates its point of articulation to the following consonant and for instance becomes a bilabial nasal before β, which itself occlusivizes to b (↗ 2.1.4.1). Further, note that a base initial r and an adjacent 1s n assimilate to the pseudo- geminate nn.¹³⁵

(393) ə-βəra → *a-n-βəra → a-m-bəra 1sS-eat.IPFV NEG-1sS-eat.IPFV ‘I eat’ ‘I do not eat’

¹³⁵Consider, however, the case of the verb ərəkkʸə ‘throw’. Although its Imperfective base begins with r it does not assimilate to the preceding nasal. Probably this r is ‘stable’ and never changes to n. Hence the unusual ə before the actual first radical r which prevents it from appearing word- initially where all r nasalize to n (↗ 2.1.3.1): (i) ə-rəkʸ → bɨ-n-rəkʸ (*bɨnnəkʸ) 1sS-throw.IPFV TEMP-1sS-throw.IPFV ‘I throw’ ‘when I throw’

139 (394) ə-βɨr → *bɨ-n-βɨr → bɨ-m-bɨr 1sS-say.IPFV TEMP-1sS-say.IPFV ‘I say’ ‘when I say’ (395) ə-rəβɨr → *a-n-rəβɨr → a-n-nəβɨr 1sS-live.IPFV NEG-1sS-live.IPFV ‘I live’ ‘I do not live’

3.11.6.4 Allomorphs of subject markers followed by object markers Some of the subject markers change their form obligatorily, some optionally, when they are followed by object markers. Refer also to section 3.12.1.1 for the ‘fused’ subject-object markers. The 1s Perfective suffix -xʷ has an allomorph -x when followed by any object marker as, for example, the 3sf primary object -na.

(396) ba-xʷ-ɨm → *ba-xʷ-na-m → ba-x-na-m say.PFV-1sS-M say.PFV-1sS-3sfO-M ‘I said’ ‘I said (to) her’

The 1p suffix -nə of the Perfective, Imperfective and the Jussive alters to -ne before most object markers, as, for example, the 2pm primary object -ku (397) or the 3sf malefactive -pa (398).

(397) nɨ-sdɨ-nə → *nɨ-sdɨ-nə-ku → nɨ-sdɨ-ne-ku 1pS-take.JUS-1pS 1pS-take.JUS-1pS-2pmO ‘let’s take’ ‘let’s take you (pm) (with us)’ (398) n-ud-nə → *n-ud-nə-pa → n-ud-ne-pa 1pS-tell.IPFV 1pS-tell.IPFV-1pS-2sfO ‘we tell’ ‘we tell to her detriment’

In contrast to this, before the primary objects featuring the consonant y, i.e. 3sf -ya, 3pm -yo and 3pf -yəma, the 1p subject marker can also be -nə.

(399) attakkən-nə-yo-m (~ attakkən-ne-yo-m) chase.away.PFV-1pS-3pmO-M ‘we chased them away’

The 3sf Perfective suffix -əc has an optional allomorph -əcə before any object marker, as, for example, the 3sm primary object -n (which triggers also labialization).

(400) bar-əc-ɨm → bʷar-əcə-n-ɨm~ bʷar-əc-n-ɨm say.PFV-3sfS-M say.PFV-3sfS-3smO-M ‘she said’ ‘she said (to) him’

The suffix -o designating 3pm in the Perfective and 2/3pm in the Imperfective

140 and Jussive is occasionally realized as -əw when followed by (and only by) the 3sm primary object -i.

(401) yɨ-βr-o → yɨ-βr-əw-i~ yɨ-βr-o-yi 3S-say.IPFV-pmS 3S-say.IPFV-pmS-3smO ‘they (m) say’ ‘they (m) say (to) him’

Note that there is either the change from -o to -əw before -i or the insertion of a glide y between -o and -i. The third conceivable output with the final 3sm -i becoming y is usually not accepted. Rather, -o-y is understood as 3pm and the purposive marker (↗ 4.7.3.6).

(402) yɨ-βr-o-y 3S-say.IPFV-pmS-PURP ‘in order that they (m) say’

3.12 Object marking 3.12.1 Forms There are three semantically distinct and mutually exclusive possibilities to mark objects on the verb by means of suffixes: ‘primary object’ (), ‘benefactive’ (), and ‘malefactive-locative-instrumental’ (). Table 61 summarizes the basic forms of these three object suffixes. Note that there are no Impersonal object suffixes. Some suffixes have allomorphs in connection with preceding subject markers (in particular ↗ 3.12.1.1). For the semantics and use of the object suffixes refer to section 3.12.2.

   ‘’ ‘’ ‘’ ‘’ ‘’ ‘’ 1s -e -n -ni -ni -βi -pi 2sm -(na)xə -kə -nxə -nkə -βxə -βkə 2sf -(na)xʸ -kʸ -nxʸ -nkʸ -βxʸ -βkʸ 3sm -n + -i -lə + -lə -wə -pʷə 3sf -na -ya -la -la -βa -pa 1p -ndə -ndə -ndə -ndə -βɨndə -pɨndə 2pm -(na)xu -ku -nxu -nku -βxu -βku 2pf -(na)xma -kma -nxɨma -nkɨma -βxɨma -βkɨma 3pm -no -yo -lo -lo -βo -po 3pf -nəma -yəma -ləma -ləma -βəma -pəma

Table 61: Object markers

The object suffixes consist in principle of a semantic ‘case’ marker plus the person- gender-number marker.  is expressed by l/n and  by β/p;  has no dedi-

141 cated marker, except for a morpheme na of the ‘light’ suffixes (see below) of the second persons in the Perfective, but not in the Imperfective an Jussive:

(403) bar-ə-naxə-m vs. yɨ-βɨr-xə vs. yə-βər-xə say.PFV-3smS-2smO-M 3smS-say.IPFV-2smO 3smS-say.JUS-2smO ‘he said you’ ‘he says you’ ‘let him say you’

For each of the three object suffixes there are two different sets, one called ‘light’ and the other one ‘heavy’ (cf. Hetzron 1977: 62).¹³⁶ The formal difference between the ‘light’ and the ‘heavy’ sets is often the strengthening of a consonant (for example .2sm -nxə vs. -nkə), but sometimes there is no difference at all (for example .3sf -la). In the case of the 3rd persons  there is a change from -n to -i/y, whereas the morpheme pair of 1s  is -e vs. -n. Also note that the 1p  and  suffixes do not differ. As for the allomorphic distribution, the ‘light’ set occurs after all singular (subject) forms except 2sf, the ‘heavy’ set after the rest, i.e. plural subjects, 2sf, and after . Table 62 shows this ‘light’-‘heavy’ distribution on the basis of one exemplary suffix (.3sm) added to the Imperfective of the verb od ‘tell’.

1s ud-wə 2sm t-ud-wə 2sf t-uj-pʷə 3sm y-ud-wə 3sf t-ud-wə 1p n-ud-ne-pʷə 2pm t-ud-o-pʷə 2pf t-ud-əma-pʷə 3pm y-ud-o-pʷə 3pf y-ud-əma-pʷə  y-uj-pʷə

Table 62:  + .3sm

The examples in (404) further show the contrast between ‘light’ and ‘heavy’ objects.

(404) ‘’ ‘’  kʼɨtʼr-e kʼɨtʼi-n kill.IMP[.2smS]-1sO kill.IMP.2sfS-1sO ‘kill meǃ’ ‘kill meǃ’

¹³⁶Hetzron (1977: 63) discusses the origin of the ‘heavy’ suffixes. According to him, they generally occur where there used to be long vowels preceding them. The loss of vowel length lead to a compensatory lengthening, i.e. gemination, of the following consonant. Subsequently, these geminates developped into the actual forms by degemination, devoicing and/or strengthening.

142  y-ud-ɨnxɨma y-uj-ɨnkɨma 3smS-tell.IPFV-BEN.2pf-M 3S-tell.IPFV.IPS-BEN.2pf-M ‘he tells for you’ ‘one tells for you’  od-ə-βo-m od-o-po-m tell.PFV-3smS-MAL.3pm-M tell.PFV-3pmS-MAL.3pm-M ‘he told to their detriment’ ‘they told to their detriment’

Labialization () as a floating feature occurs with the light object markers of 3sm  and 3sm .  affects the rightmost labializable consonant (labials and velars), very similar to the labialization in the formation of the Impersonal as described in section ↗ 3.11.5. In the suffixes of 3sm , labialization is also present but it affects the labializable ʽcaseʼ marker (β→w / p→pʷ). Further, note that the subject marker 1s -xʷ of the Perfective is not labialized when followed by object suffixes (↗ 3.11.6), but being the rightmost labializable consonant it becomes the host for the floating . The examples in (405) illustrate labialization with 3sm  and , in comparison to the forms with the feminine object markers.

(405) a. bʷar-ə-n-ɨm vs. bar-ə-na-m say.PFV-3smS-3smO-M say.PFV-3smS-3sfO-M ʽhe said himʼ ʽhe said herʼ b. ba-xʷ-ɨn-ɨm vs. ba-x-na-m say.PFV-1sS-3smO-M say.PFV-1sS-3sfO-M ʽhe said himʼ ʽhe said herʼ c. yɨ-xʷəl-lə vs. yɨ-xəl-la 3smS-become.IPFV-BEN.3sm 3smS-become.IPFV-BEN.3sf ʽit is possible for himʼ ʽit is possible for herʼ

3.12.1.1 Fused subject-object markers As can be observed in the examples above, the object markers are suffixed directly to the conjugated verb forms.¹³⁷ This means that in the Perfective they are adjacent to the subject markers and in the Imperfective and Jussive they follow directly the verb base in the singular on the on hand or the number/gender subject suffixes in the plural on the other hand. In some instances, as described below, the adjacent subject and object markers fuse to an inseparable portmanteau morpheme. In the Perfective, the 1s marker -x- and the element -x- of the second persons

¹³⁷Further markers as the main verb marker -m (↗ 3.18.1.1), the futures -ʃə and -te (↗ 3.18.5), or subordinators like -e and -xəma (↗ 4.7) always appear at the very end of the verb form following the object suffixes: (i) yɨ-wəsd-ɨβi-te (ii) tɨ-ʃə-n-xəma 3sm-take.-.1s-. 3sm-want.-3sm- ‘he will take to my detriment / from me’ ‘that she wants it’

143 of any of the object markers fuse to an inseparable -k. Remember that 1s is not labialized (*-xʷ-) in combination with object markers (↗ 3.11.6). The examples in (406) illustrate the fused forms of 1s and the primary object. Note that the palatalization of 2sfO -xʸ is also present in the fused -kʸ (b).

(406) a. *od-x-xə-m → od-kə-m tell.PFV-1sS-2smO-M tell.PFV-1sS.2smO-M ‘I told you’ b. *od-x-xʸ-m → od-kʸ-ɨm tell.PFV-1sS-2sfO-M tell.PFV-1sS.2sfO-M ‘I told you’ c. *od-x-xu-m → od-ku-m tell.PFV-1sS-2pmO-M tell.PFV-1sS.2pmO-M ‘I told you’ d. *od-x-xəma-m → od-kəma-m tell.PFV-1sS-2pfO-M tell.PFV-1sS.2pfO-M ‘I told you’

The fusion of two x into k as shown in (406) above is reminiscent of gemination of x, which results in (k)k in almost all cases. Consider, for example, the geminated Perfective səkkər-ə vs. the non-geminated Imperfective yɨ-səxɨr of the verb √sxr ‘get drunk’. Yet, the fused subject-object markers present a different case that cannot readily be set on the same level with ‘real’ gemination for two reasons. Firstly, the combination x+x across morpheme boundaries does not yield k elsewhere. This concerns in particular Perfective verbs with a base final x and 1s -xʷ (407) or any other subject suffix beginning with x.

(407) manəx-xʷ-ɨm (*manəkkʷɨm,*manəkʷɨm) capture.PFV-1sS-M ‘I captured’

Secondly, the merging of the two x to k also occurs when they are not adjacent as it is the case with benefactive (408) and malefactive (409) objects. Even though the position of the respective ‘case’ markers n and β is between the subject and the object person marker (i.e. with the other, nonfusing persons), there is only one fused k for both persons, remarkably after the ‘case’ marker.

(408) a. *od-x-nxə-m → od-ɨnkə-m tell.PFV-1sS-BEN.2sm-M tell.PFV-1sS.BEN.2sm-M ‘I told for you’ b. *od-x-nxʸ-m → od-ɨnkʸ-ɨm tell.PFV-1sS-BEN.2sf-M tell.PFV-1sS.BEN.2sf-M ‘I told for you’

144 c. *od-x-nxu-m → od-ɨnku-m tell.PFV-1sS-BEN.2pm-M tell.PFV-1sS.BEN.2pm-M ‘I told for you’ d. *od-x-nxəma-m → od-ɨnkəma-m tell.PFV-1sS-BEN.2pf-M tell.PFV-1sS.BEN.2pf-M ‘I told for you’ (409) a. *od-x-βxə-m → od-ɨβkə-m tell.PFV-1sS-MAL.2sm-M tell.PFV-1sS.MAL.2sm-M ‘I told to your detriment’ b. *od-x-βxʸ-m → od-ɨβkʸ-ɨm tell.PFV-1sS-MAL.2sf-M tell.PFV-1sS.MAL.2sf-M ‘I told to your detriment’ c. *od-x-βxu-m → od-ɨβku-m tell.PFV-1sS-MAL.2pm-M tell.PFV-1sS.MAL.2pm-M ‘I told to your detriment’ d. *od-x-βxəma-m → od-ɨβkəma-m tell.PFV-1sS-MAL.2pf-M tell.PFV-1sS.MAL.2pf-M ‘I told to your detriment’

Nevertheless, occasionally one can also hear a geminated fused kk (410a). It seems that this is only possible with verbs that have a weak final radical whereas sound verbs never showing gemination (410b).

(410) a. cʼənə-kkə-m ~cʼənə-kə-m give.birth.PFV-1sS.2smS-M ‘I gave birth to you (sm)’ b. an-nəgəd-kə (*an-nəgəd-ɨkkə) NEG-touch.PFV-1sS.2smO-M ‘I did not touch you’

There are some further cases of ‘fusion’ between subject and object suffixes. In the Perfective the 1s primary object -e deletes the preceding ə of 2sm -xə and 3sm -ə (411). In the latter case (411b) this results in the complete disappearance of the subject marker since it consists only of -ə.

(411) a. *od-xə-e-m → od-xe-m tell.PFV-2smS-1sO-M tell.PFV-2smS.1sO-M ‘you (sm) told me’ b. *od-ə-e-m → od-e-m tell.PFV-3smS-1sO-M tell.PFV-3smS.1sO-M ‘he told me’

The 3sm primary object -i fuses with the 1p subject suffix -nə to -ne in the Per-

145 fective (412a), Imperfective (412b) and Jussive (412c). Note that the allomorph of -nə before object markers generally is -ne (↗ 3.11.6.4), thus one could also argue that 3smS -i is deleted after -ne.

(412) a. *xəna-nə-i-m → xəna-ne-m prevent.PFV-1pS-3smO-M prevent.PFV-1pS.3smO-M ‘we prevented him’ b. *nɨ-tkʼʸəppən-nə-i → nɨ-tkʼʸəppən-ne 1pS-accept.IPFV-1pS-3smO 1pS-accept.IPFV-1pS.3smO ‘we accept him’ c. *nɨ-mʷəkkɨn-nə-i → nɨ-mʷəkkɨn-ne 1pS-try.JUS-1pS-3smO 1pS-try.JUS-1pS.3smO ‘let’s try him’

Further, in all of the three basic TAM forms Perfective (413), Imperfective (414) and Jussive (415) the ‘heavy’ 3sm primary object -i can fuse with the preceding feminine plural markers -xma and -əma to -xmɛ and -əmɛ respectively.¹³⁸ How- ever, this is only a dialectal variant that is more frequent in villages close to the Chaha area. The pronunciation with the diphthong, i.e. a+i→ay, is more typical of Gumer.

(413) a. *od-xɨma-i-m → od-xɨmɛ-m ~ od-xɨma-y-ɨm tell.PFV-2pfS-3smO-M tell.PFV-3pfS.3smO-M tell.PFV-2pfS-3smO-M ‘you (p) told him’ b. *od-əma-i-m → od-əmɛ-m ~ od-əma-y-ɨm tell.PFV-3pfS-3smO-M tell.PFV-3pfS.3smO-M tell.PFV-3pfS-3smO-M ‘they () told him’

(414) a. *t-ud-əma-i → t-ud-əmɛ ~ t-ud-əma-y 2S-tell.IPFV-pfS-3smO 2S-tell.IPFV-pfS.3smO 2S-tell.IPFV-pfS-3smO ‘you (p) tell him’ b. *y-ud-əma-i → y-ud-əmɛ ~ y-ud-əma-y 3S-tell.IPFV-pfS-3smO 3S-tell.IPFV-pfS.3smO 3S-tell.IPFV-pfS-3smO ‘they () tell him’

(415) a. *od-əma-i → od-əmɛ ~ od-əma-y tell.IMP-2pfS-3smO tell.IMP-2pfS.3smO tell.IMP-2pfS-3smO ‘tell (p) him!’ b. *y-od-əma-i → y-od-əmɛ ~ y-od-əma-y 3S-tell.JUS-pfS-3smO 3S-tell.JUS-pfS.3smO 3S-tell.JUS-pfS-3smO ‘let them (p) tell him!’

¹³⁸Often the realization is -xme and -əme with a quite high e rather than ɛ (also ↗ 2.2.1).

146 Finally, note that 3sm -i after 2pm -xu and 2/3pm -o usually cannot form a diphthong but needs an epenthetic glide y (416)-(417). In the latter case, the diphthong -oy is mostly only understood as the subject marker followed by the purposive marker (418) (↗ 4.7.3.6).

(416) od-xu-yi tell.PFV-2pmS-3smO ‘you (pm) told him’ (417) y-ud-o-yi 3S-tell.IPFV-pmS-3smO ‘they (m) told him’ (418) y-ud-o-y 3S-tell.IPFV-pmS-PURP ‘in order for them (m) to tell’

3.12.2 Use of the object markers As mentioned earlier, the object markers are mutually exclusive. Any combinations of two or more suffixes on one verb are impossible, even when a state of affairs involves more than one non-subject participant:

(419) *od-ə-n-lə-m tell.PFV-3smS-3smO-BEN.3sm-M (intended: ‘He told [it] to him for him.’)

In such cases one suffix has to “win” over the other. Usually the competing pairs are primary object vs. benefactive and primary object vs. malefactive (and virtually never benefactive vs. malefactive). There are many factors that play a role in the choice of the right object marker. As a general rule, in sentences that require a definite object and a beneficiary or maleficiary, the latter two seem to rule out the primary object in most cases:

(420) a. astəmari-xʷɨt nəkəβ-o-yi-m. teacher-DEF.sm find.PFV-3pmS-3smO-M ‘They found the teacher.’ b. astəmari-xʷɨt nəkəβ-o-lə-m. teacher-DEF.sm find.PFV-3pmS-BEN.3sm-M ‘They found the teacher for him.’ c. astəmari-xʷɨt nəkəβ-o-pʷə-m. teacher-DEF.sm find.PFV-3pmS-MAL.3sm-M ‘They found the teacher to his detriment.’

However, while the overt marking of the beneficiary as in (421) is the norm, it is not ruled out to have the primary object on the verb as in (422). It seems that factors like information structure play a role here, but the exact conditions are yet to be explored.

147 (421) tʼay-xʷɨta y-ərc-əta sɨyə-lə-m. sheep-DEF.sm DAT-son-3smPOSS buy.PFV[.3smS]-BEN.3sm-M ‘He bought the sheep for his son.’ (422) tʼay-xʷɨta y-ərc-əta sɨyə-n-ɨm. sheep-DEF.sm DAT-son-3smPOSS buy.PFV[.3smS]-3smO-M ‘He bought the sheep for his son.’

3.12.2.1 Primary object The primary object suffix marks direct objects of transitive verbs (423)-(424), recipients (425) and addressees (426) of ditransitive verbs, and experiencers of experiencer verbs (427).

(423) amədar kʼʷəttʼər-ə-n-ɨm. cold kill.PFV-3smS-3smO-M ‘The cold (weather) killed him.’ (424) nɨkʼyə gʷəppe-yəna aʃʃə-xʷ-ɨn-ɨm. big brother-1sPOSS see.PFV-1sS-3smO-M ‘I saw my older brother.’ (425) mɨkʼar aβ-o-ku-m? what give.PFV-3pmS-2pmO-M ‘What did they give you (pm)?’ (426) ɨyya ə-tot-ɨn-ʃə ba-x-na-m. 1s 1sS-make.JUS-3smO-FUT.INDEF say.PFV-1sS-3sfO-M ‘I said to her “I will make it”.’ (427) tɨkə-na gʷad-ə-n-ɨm. child-1sPOSS be.hungry.PFV-3smS-3smO-M ‘My child is hungry.’

The occurrence of the primary object suffix on the verb is primarily conditioned by the definiteness of its referent. While indefinite or generic objects cannot appear on the verb (428), definite ones often are marked (429a), but as example (429b) shows not necessarily.

(428) bora antʼɨ-nə-m. ox cut.PFV-1pS-M ‘We slaughtered an ox.’ (429) a. bet-xʷɨta nəkəβ-o-yi-m. house-DEF.sm find.PFV-3pmS-3smO-M ‘They found the house.’ b. bet-xʷɨta nəkəβ-o-m. house-DEF.sm find.PFV-3pmS-M ‘They found the house.’

148 Definiteness is by far not the only factor that plays a role. Apparently, there is a complicated interaction between multiple other factors such as specificity, topicality, discourse pragmatics, animacy, etc. A solid and detailed description of the occurrence of the primary object suffixes requires a large corpus and is beyond the scope of this thesis.

3.12.2.2 Benefactive and malefactive The benefactive and malefactive suffixes mark, as their labels suggest, beneficiaries (430) and maleficiaries (431).

(430) tʼay antʼ-ɨxʷ-ɨlə-m. sheep cut.PFV-1sS-BEN.3sm-M ‘I slaughtered a sheep for him.’ (431) gʷəncə cənə-βi-m. hyena come.PFV[.3smS]-MAL.1s-M ‘A hyena came to me (and that is dangerous).’

The latter, however, also has a local and an instrumental meaning corresponding to the (cognate) prefix bə- (↗ 4.7.1.4). Thus, without context the meaning of a malefactive suffix can be ambiguous (432).

(432) kʼɨtʼɨr-wə! kill.IMP[.2smS]-MAL.3sm ‘Kill [it] in it!’ ~ ‘Kill [it] with it!’~ ‘Kill [it] to his detriment!’

A concise but detailed description of the use of beneficiaries and maleficiaries is found in Völlmin (2010a).

3.13 Formation of the Infinitive The Infinitive (or Verbal Noun) in Gumer can be formed in two different ways, either with the prefix wə- or the suffix -ot. Both of them attach to the Jussive base, which among other things means that type B verbs feature depalatalization also in the Infinitives. In the case of verbs with final I or U, the Infinitive with -ot also uses the depalatalized Jussive base while the Infinitive with wə- exhibits the palatalized form. The practical rule of thumb is as follows: Infinitives with wə- look like the Jussive 3sm with w- instead of y-, for example yəzɨmd → wəzɨmd ‘pull’ and Infinitives with -ot look like the Imperative 3pm with an additional t, for example zəngo → zəngot ‘speak’. Table 63 lists both Infinitives of all basic sound verbs. Table 64 presents the Infinitives of those weak verb types that feature depalatalization in combination with -ot. Note that these depalatalized forms generally can have two variants, i.e. with or without additional -əβ (like 2pm and 3pm). Examples are bɨxot ~bɨxəβot √βxI ‘cry, gɨrot ~gɨrəβot √grI ‘cast a spell’ or tʼɨwot ~tʼɨwəβot √tʼβU ‘suck’.

149 Type Root  wə-  -ot

A1 √skʼr wəskʼɨr sɨkʼrot ‘hang up’ A2 √rtʼr wəntʼər nɨtʼərot ‘melt ()’ B √zβr wəzəppɨr zəprot ‘return ()’ C √ʒrg wəʒarg ʒargot ‘go away’ D √brs wəwəns bʷənsot ‘feel lonely’ E √grdm wəgərdɨm gərdɨmot ‘break s.th. in two’ F √frtx wəfratɨx fɨratxot ‘mess’

Table 63: Infinitives of sound verbs

Type Root  wə-  -ot

A1-12U √kʼrU wəkʼʷi kʼʷɨrot ‘roast’ A1-12I-y √ftʼI wəfcʼ fɨtʼot ‘sharpen; grind’ A2-12I √stʼI wəstʼe sɨtʼəβot ‘drink’ C-12U √ftʼU wəfʷacʼ fʷatʼot ‘mow’ C-12I √kʼsI wəkʼaʃ kʼasot ‘throw away’ E-123U √kʼrtʼU wəkʼʷəncʼ kʼʷəntʼot ‘take a handful’ E-123I √zrgI wəzəngʸ zəngot ‘speak’ F-123I √βrtʼI wəβracʼ bɨratʼot ‘scatter’

Table 64: Infinitives of (depalatalizing) weak verbs

Derived stems form the Infinitives analogously as the selected examples in table 65 show. In particular notice again the depalatalization in the first two verbs derived from a type B root and a root with final I. Further, the morphophonological rules are the same as elsewhere: the vowel ə of the prefix wə-, which precedes the derivational affixes, is deleted by following a- or at-, and the suffix -ot, which follows the base, deletes a final a; the morpheme tə- is reduced to t- when not standing word-initially; and finally gemination with base final r is not present when a morpheme (i.e. here -ot) follows (↗ 3.3.2).

150 Stem Type Root  wə-  -ot tə- B √kʼrβ wətkʼənəβ təkʼənəβot ‘be near’ E √zrgI wətzangʸ təzangot ‘talk (together)’ a- A √βrA waβra aβrot ‘feed’ A √xdr waxdɨr axdɨrot ‘dress’ at- A √gfr watgəffɨr atgəfrot ‘free’ A √rgd watragd atragdot ‘cause to touch one another’

Table 65: Infinitives of (selected) derived stems

The Infinitive with -ot can be negated by the negation marker an- (433).

(433) dəpr-ot → an-dəpr-ot finish-INF NEG-finish-INF ‘finish’ ‘not finish’

Note that tə- becomes t- word-internally (434) (↗ 3.6.1.1).

(434) təfəkkʼər-ot məβt-əna-w. → an-tɨfəkkʼər-ot məβt-əna-w. play-INF right-1sPOSS-COP.3smS NEG-play-INF right-1sPOSS-COP.3smS ‘It is my right to play.’ ‘It is my right not to play.’

In contrast, it is not possible to form negated Infinitives with wə-, for instance *w-an-dəppɨr, *w-an-tɨfəkkʼər.

3.14 Formation and uses of the converbs The converbs in Gumer and their formation and functions are discussed in detail in Völlmin (2010b). The following sections is a summary of the main points.

3.14.1 Formation of the t-converb The so-called t-converb (.) is a remarkable verb form both formally and functionally. It is named after its characteristic morpheme -t- (or -tə-), which is added to a form that looks exactly like (but neither is derived from nor corresponds functionally to) the feminine singular Imperative, or in other words the Jussive base with palatalization. Consider the illustrative examples in (435).

(435)   + . √tʼβtʼ tʼəβətʼ tʼɨβtʼ tʼɨβcʼ tʼɨβcʼ-ɨt- ‘take, grasp, hold’ √srkʼ sənəkʼ sɨrkʼ sɨrkʼʸ sɨrkʼʸ-ɨt- ‘steal’ √Ur wər wər we we-t- ‘go’ √tIrI cənə tən tən tən-t- ‘come’

151 According to most authors (Hetzron 1977: 96; Rose 2007: 416; Leslau 1983: 19; Leslau 1992: 444) the morpheme of the t-converb is -tə-. However, in Gumer I have mostly encountered a variant -t- without vowel ə (cf. also Polotsky 1951: 45). It is not clear if there are indeed two coexisting variants or if this is a case of dialectal variation between Gumer and Chaha (or alternatively this could even be a mistake in that the 3sm, which ends in tə, was analyzed as the default form). The t-converb is fully conjugated. As shown in table 66 with the verb tʼəβətʼ ‘take, grasp, hold’, the subject markers suffixed to -t- correspond to the ones found with the Perfective (↗ 3.11.1). As for the Impersonal, however, there is one striking difference: the t-converb features a subject suffix -o which looks like (or is the same) as 3pm, in contrast to all other Impersonals that are formed distinctively and do not have, apart from the dummy (third person) subject prefix y- or yə- in the Imperfective and Jussive, such a subject marking.

  1 tʼɨβcʼɨ-t-ɨxʷ tʼɨβcʼɨ-t-nə 2m tʼɨβcʼɨ-t-xə tʼɨβcʼɨ-t-xu 2f tʼɨβcʼɨ-t-xʸ tʼɨβcʼɨ-t-xɨma 3m tʼɨβcʼɨ-t-ə tʼɨβcʼɨ-t-o 3f tʼɨβcʼɨ-t-əc tʼɨβcʼɨ-t-əma  tʼɨβcʼɨ-t-o

Table 66: t-converb of √tʼβtʼ ‘take, grab, hold’

The t-converb can further be furnished with object markers and the linker -ta(nə) (↗ 3.14.4). Both are found on the same t-converb in example (436).

(436) yə-xno bəz-ɨm tɛ-tə-x-no-tanə fərəz-əna tʼɨβcʼɨ-t-ɨxʷ DAT-3pm here-ALSO leave-CV.T-1sS-3pmO-LINK horse-1sPOSS take-CV.T-1sS wə-kra nər-ə-βi ba-xʷ-ɨm wəssən-xʷ-ɨm. INF-ascend EX-3smS-MAL.1s say.PFV-1sS-CV.M decideѦ.PFV-1sS-M ‘I decided that I have to go up, leaving them here and taking my horse.’

Use and distribution of the t-converb in contrast to the m-converb are discussed in section 3.14.3.

3.14.2 Formation of the m-converb In contrast to the t-converb, the so-called m-converb is not a single dedicated verb form. Rather it is formed by the suffix -m (.) added to any (matrix) verb form as can be seen in (437) (repeated from Völlmin 2010b: 85). Moreover, this also includes infinitives (i.e. verbal nouns) (cf. Hetzron 1977: 94), a fact which points toward a common origin of converbal and coordinating -m.

152 (437) matrix verb .  səppər-ə-m səppər-ə-m  yɨ-səβɨr yɨ-səβr-ɨm  yə-sβɨr yə-sβɨr-ɨm  sɨβɨr sɨβr-ɨm

The by far most frequent m-converb is with perfective forms. Due to the fact that matrix (‘sentence-final’) perfective verbs obligatorily feature the formally identical main verb marker -m (↗ 3.18.1.1), perfective m-converbs are not distinguishable from perfective matrix verbs, as in (438)-(439). Nevertheless, they are functionally distinguishable (↗ 3.14.5).

(438) tʼəβətʼ-xʷ-ɨm wər-xʷ-ɨm. take.PFV-1sS-CV.M go.PFV-1sS-M. ‘I took [it] with me.’ (more lit.: ‘Having taken [it], I went.’ or ‘I took [it] and went.’) (439) at gənə wər-o-m attər-əβo-m. one country go.PFV-3pmS-CV.M spend.night-3pmS-M ‘They went to a country and spent the night [there].’

Generally speaking, m-converbs of ,  and  tend to occur in chaining events and have to be followed by a matrix verb with the same TAM form, as shown in examples (440)-(441).

(440) yə-mʷəxɨr yɨ-rəx-ɨm mʷəxɨr yɨ-cəl-lə-βa. DAT-Muher 3smS-send.IPFV-CV.M Muher 3smS-come.IPFV-BEN.3sm-AUX.PT ‘He would send for the Muher and the Muher would come to [help] him.’

(441) ʃərət bɨra-m wɨtʼa! foot eat.IMP[.2smS] go.out.IMP[.2smS] ‘Eat the food and leave!’

In contrast, non-perfective matrix verbs can be (and often are) preceded by a perfective m-converb (442).

(442) bora antʼ-ɨnə-m nɨ-ʃəd-nə. ox cut.PFV-1pS-CV.M 1pS-distribute.IPFV-1pS ‘We slaughter an ox and distribute [it].’

3.14.3 m-converb vs. t-converb The m-converb and the t-converb do not express specialized or specific meanings. Rather they are distributed complementary: the t-converb occurs before verb forms or states of affairs that can be subsumed under the label ‘irrealis’. This includes in particular negated verbs (443) and (indefinite) futures (444), but also contexts such as necessity (445) (see Völlmin 2010b: 86).

153 (443) tʼɨβcʼ-ɨt-ɨxʷ an-wər-xʷ. take-CV.T-1sS NEG-go.PFV-1sS ‘I did not take [it] with me.’ (444) tʼɨβcʼ-ɨt-ɨxʷ ə-wər-ʃə. take-CV.T-1sS 1sS-go.JUS-FUT.INDEF ‘I will not take [it] with me.’ (445) tʼɨβcʼ-ɨt-ɨxʷ wər-ot nər-ə-βi. take-CV.T-1sS JUS-INF EX-3smS-MAL.1s ‘I have to take [it] with me.’

In all other (i.e. ‘realisʼ) contexts the m-converb is used as in example (438) above. Note that the definite future (↗ 3.18.5) usually occurs with the m-converb (446).

(446) tʼəβətʼ-xʷ-ɨm ar-te. take.PFV-1sS-CV.M [1sS.]go.IPFV-FUT.DEF ‘I will take [it] with me.’

3.14.4 The converbal linker -tanə~ -ta Both m-converb and t-converb can be suffixed with an additional linking element -tanə or its short form -ta (447)-(448).

(447) fərəz wəsəd-xɨ-wə-m-tanə gʸɨβat gəppa-xʷ-ɨm. horse take.PFV-1sS-MAL.3s-CV.M-LINK riding enter.PFV-1sS-M ‘I took a/the horse from him and started riding.’ (448) we-t-nə-tanə a-m-bʷəga-ne. go-CV.T-1pS-LINK NEG-1pS-fight.IPFV-1pS.3smO ‘We do not go and attack him [first].’

The possibility of adding -tanə ~-ta to a (perfective) verb form allows to distinguish between m-converb and (perfective) matrix verb. In (449), -ta on notʼəm is optional, however the fact that it can be used there shows the verbʼs status as converb.

(449) tʼay-xʷɨt gʷəncə b-i-yaʒ notʼ-ə-m(-ta) wədərə sheep-DEF.m hyena TEMP-3smS-see.IPFV run.PFV-3smS-CV.M(-LINK) rope metʼəs-ə-m. detach.PFV-3smS-M ‘When the sheep saw the hyena, it ran and detached the rope.’

It has been suggested that -tanə ~-ta is obligatory in sequences of clearly separate events (expressing ‘and then’) and optional in all other functions (see Hetzron 1977: 97) (↗ 3.14.5). Nevertheless, arguably it is probably more accurate to state that -tanə ~ -ta is required in case the converb could also be understood as an adverbial modifier (↗ 3.14.5) of the matrix verb (rather than a separate preceding event, cf. Völlmin 2010b: 92). In contrast, the presence of -ta (but usually not -tanə) is obligatory when the converb follows the matrix verb as an afterthought. The

154 additional verb tʼəmmen-ta ʽbeing thirsty, because I was thirstyʼ in (450) without -ta could only be understood as an independent declarative utterance meaning ‘I was thirsty’.

(450) bə-dərar ambo ɨxa acənə-xʷ-ɨm səccʼə-xʷ-ɨm, LOC-back.part.of.house A. water bring.PFV-1s-CV.M drink.PFV-1sS-M tʼəmm-e-n-ta. be.thirsty.PFV-3smS.1sO-M.CV-LINK ‘I brought an Ambo water (i.e. bottle of mineral water) from the dərar and drank it, because I was thirsty (me being thirsty).’

3.14.5 Functions of the converbs The various functions of the converbs are discussed in detail in Völlmin (2010b). In summary, two basic functions can be distinguished: (general) chaining and (adverbial) modification. Chaining is the semantically loose connection of a series of two or more events, ranging from ‘pureʼ chains of subsequent separate events (451)¹³⁹, to connections of (two) events that belong closely together describing one conceptual unit (452). Note that in the latter example the verbs share the same object.

(451) təxankʼʸətə biro wəsəd-o-m fɨraʃ acənə-βo-m afterwards officeѦ take.PFV-3pmS-CV.M mattressѦ bring.PFV-3pmS-CV.M bə-fraʃ attən-nə-m. LOC-mattress spend.the.night.PFV-1pS-M ‘Then they took [us] to the office, prepared mattresses, we spent the night on the mattresses.’ (452) kʷɨtara antʼ-ə-m ʃəkət-ə-m. chicken cut.PFV-3smS-CV.M prepare.PFV-3smS-M ʽHe slaughered and prepared a chicken.ʼ

In contrast, (adverbial) modification stands for events that are semantically depen- dend on the matrix verb, often corresponding to subordinate clauses or adverbs in European languages. The converbal part hereby expresses accompanying activities (453) or the manner of the matrix event (454)-(455).

(453) kʼəya-ta t-i-səra gamʷə yɨlɨl bʷar-i-m village-3smPOSS TEMP-3smS-arrive.IPFV time cheering say.PFV.IPS-3smO-CV.M andɨr dənəgʷ-i-m yɨ-tkʼʸəppʷər-i. drum hit.PFV.IPS-3smO-CV.M 3-welcome.IPFV.IPS-3smO ʽWhen he arrives in his village they welcome him by cheering and hitting drums.ʼ

¹³⁹Admittedly, such a sequence of verbs could also be understood as three independent sentences a perfective matrix verb each.

155 (454) afəttʼər-xə-m cənə-xə-m? hurry.PFV-2smS-CV.M come.PFV-2smS-M ʽDid you come quickly?ʼ (455) kʷɨtara-xʷɨt bənnər-ə-m wər-ə-m. chicken-DEF.m fly.PFV-3smS-M.CV go.PFV-3smS-M ʽThe chicken flew away.ʼ

There are further grammaticalizations of the converb. On the one hand, it is used in the periphrastic construction . + xar ʽknowʼ to express an experiential perfect (456).

(456) amerika wər-xu-m tɨ-xʸr-o? A. go.PFV-2pmS-CV.M 2S-know.IPFV-pmS ʽHave you (ever) been to America?ʼ

On the other hand some verbs occur as adpositions (457), and finally the converb of bar ʽsayʼ as complementizer (↗ 3.17.2).

(457) tə-trama kʼənəs-ə-m-ta ABL-yesterday begin.PFV-3smS-CV.M-LINK ʽstarting from yesterdayʼ

Finally, note that converbs cannot be negated. Instead temporal clauses are used (↗ 4.7.3.3).

3.15 Verbal negation Verbs are negated by a negation marker prefixed to the conjugated verb. For the Perfective the marker is an-, for the Imperfective and Jussive a-, and for the Pro- hibitive ɨn-.

3.15.1 Negated Perfective As shown in table 67, the Perfective is negated with an- directly prefixed to the base.

  1 an-kəfət-xʷ an-kəfət-nə 2m an-kəfət-xə an-kəfət-xu 2f an-kəfət-xʸ an-kəfət-xɨma 3m an-kəfət-ə an-kəfət-o 3f an-kəfət-əc an-kəfət-əma  an-kəfʷəc(-i)

Table 67: Negated  of √kft ‘open’

156 The conjugation remains unaffected,¹⁴⁰ but the main verb marker -m, which is obligatory in main clauses, is dropped (458) (↗ 3.18.1.1). Thus, the negated forms can constitute a full predication (other than the bare Perfective in table 56 above).

(458) aʃʃə-xʷ-ɨm. vs. an-aʃʃə-xʷ. see.PFV-1sS-M NEG-see.PFV-1sS-M ‘I saw.’ ‘I did not see.’

The nasal n of the negation marker assimilates to the point of articulation of the following consonant (↗ 2.1.4.1), for example an-kəfət-ə → [aŋkəfətə] ‘he did not open’ or an-mʷət-ə → [ammʷətə] ‘he did not die’. When an- meets w there are two possible outputs, for example an-wəttʼa → [aŋwəttʼa] or [ambʷəttʼa]. Quite often the n nasalizes the a and even completely disappears, especially with velars, resulting in variants [aŋkəfətə]~ [ãŋkəfətə] ~ [ãkəfətə] (cf. example (49e)). Finally, I have recorded an instance of a spontaneous Amharic loan beginning with l, to which the nasal of the negation fully assimilated, i.e. an-ləffa-nə → [alləffanə] ‘we did not get tired’.

3.15.2 Negated Imperfective As illustrated in table 68, the Imperfective is negated with a-. It is directly prefixed to the subject markers and has an influence on the shape of some of them. For one thing, the ə- of the first person singular is changed to -n-, adjusting the form to the first person prefix elsewhere and thus leveling the singular-plural difference of the ordinary Imperfective. For another thing, the negation marker a- fuses with the subject prefix y- of the third persons yielding a portmanteau morpheme e-. This e- can sometimes be realized more open similar to [ɛ] (compare the occasional realization of word-final ay as ɛ, ↗ 2.2.1).

  1 a-n-kəft a-n-kəft-ɨnə 2m a-t-kəft a-t-kəft-o 2f a-t-kəfc a-t-kəft-əma 3m e-kəft e-kəft-o 3f a-t-kəft e-kəft-əma  e-kəfʷc(-i)

Table 68: Negated  of √kft ‘open’

Note the different outcome of a root initial r in the affirmative and negative first person singular Imperfective (459). Following the general allophony rules of r/n

¹⁴⁰In some Gurage varieties such as Muher or Ezha negated Perfectives lose their gemination (cf., for example, Hetzron 1977: 87f.), but in Gumer the bases do not change when negated.

157 (↗ 2.1.3), it is realized as r after the vocalic subject marker ə, but it assimilates to the directly preceding n of the negated Imperfective.

(459) ə-rəmd. vs. a-n-nəmd. 1sS-love.IPFV NEG-1sS-love.IPFV ‘I love.’ ‘I do not love.’

Negated Imperfectives are not only the negative forms of Imperfectives, but also negate the two Futures (460) (↗ 3.18.5).

(460) yɨ-cən.  3smS-come.IPFV   ‘He comes.’    e-cən.   NEG.3smS-come.IPFV yɨ-cən-te.   3smS-come.IPFV-FUT.DEF  ‘He does not come.’ ‘He will (certainly) come.’ ‘He will (certainly) not come.’   ‘He will (probably) not come.’  yɨ-tən-ʃə.   3smS-come.JUS-FUT.INDEF    ‘He will (probably) come.’   3.15.3 Negated Jussive As shown in table 69, the Jussive is negated with a-. The subject affixes are not the same as in the affirmative Jussive but correspond to the ones of the (negated) Imperfective (↗ 3.15.2). In particular there is no special form (yə-) for the third persons – rather, the prefixed a- fuses with the initial glide y- to e- like in the Imperfective – and second persons appear with the subject marker t- in contrast to their affirmative counterparts, the Imperatives.

  1 a-n-kɨft a-n-kɨft-ɨnə 2m a-t-kɨft a-t-kɨft-o 2f a-t-kɨfc a-t-kɨft-əma 3m e-kɨft e-kɨft-o 3f a-t-kɨft e-kɨft-əma  e-kɨfʷc(-i)

Table 69: Negated  of √kft ‘open’

158 3.15.4 Prohibitive The Prohibitive is built with the negation prefix ɨn- and the inflected forms of the Perfective, illustrated in table 70.

  1 ɨn-kəfət-xʷ ɨn-kəfət-nə 2m ɨn-kəfət-xə ɨn-kəfət-xu 2f ɨn-kəfət-xʸ ɨn-kəfət-xɨma 3m ɨn-kəfət-ə ɨn-kəfət-o 3f ɨn-kəfət-əc ɨn-kəfət-əma  ɨn-kəfʷəc(-i)

Table 70: Prohibitive of √kft ‘open’

Note again that the nasal n of the negation marker assimilates to the point of articulation of the following consonant, for example ɨn-kəfət-xə → [ɨŋkəfətxə] ‘do not openǃ’ or ɨn-βəna-xu → [ɨmbənaxu] ‘do not eatǃ’ (↗ 2.1.4.1).

3.15.4.1 Negated Jussive vs. Prohibitive The negated Jussive and the Prohibitive are similar in meaning and often interchangeable. As the label suggests, however, the Prohibitive seems to express a stronger prohibition or interdiction than the negated Jussive. The latter, on the other hand, tends to be used to signify a wish or hope rather than a strict order. The exact interpretation or translation may also differ depending on the grammatical person. As for the second persons, which occur more frequently than the first or third persons, both forms are used to negate the Imperative. Formally, there is no negated Imperative, i.e. there is no negated second person Jussive without the subject prefix. Thus instead of *a-kɨft one has to say a-t-kɨft () or ɨn-kəfət-xə () ‘do not open!’. As mentioned above, these two possibilities appear to be mutually interchangeable without significant differences in meaning.

3.15.5 Negation of Past Imperfective ( + banə) The Past Imperfective has a special way of forming the negation. The past auxiliary banə ~ -βa follows (as expected) the conjugated Imperfective in the affirmative (461a), but when negated it appears as prefix b- directly attached to the negated Imperfective (461b).¹⁴¹ Table 71 shows the full paradigm.

¹⁴¹Even though it seems that banə (and its shorter form -βa) is connected with b-, this does not mean that it is banə itself that is prefixed.

159 (461) a. yɨ-cən banə 3smS-come.IPFV AUX.PT ‘he used to come’ b. b-e-cən AUX.PT-NEG.3smS-come.IPFV ‘he did not use to come’

  1 b-a-n-kəft b-a-n-kəft-ɨnə 2m b-a-t-kəft b-a-t-kəft-o 2f b-a-t-kəfc b-a-t-kəft-əma 3m b-e-kəft b-e-kəft-o 3f b-a-t-kəft b-e-kəft-əma  b-e-kəfʷc(-i)

Table 71: Negated Past Imperfective of √kft ‘open’

3.15.6 Negation of Past Perfective ( + banə) Other than the Past Imperfective, the Past Perfective (462a) is negated regularly like the simple Perfective with prefixed an- and dropping of the main verb marker -m, followed by the invariable past auxiliary ba(nə) (462b). Table 72 shows the full paradigm.

(462) a. təkʼaw-xʷ-ɨm banə drink.coffee.PFV-1sS-M AUX.PT ‘I had drunk coffee.’ b. an-tɨkʼaw-xʷ banə NEG-drink.coffee.PFV-1sS AUX.PT ‘I had not drunk coffee.’

  1 an-kəfət-xʷ banə an-kəfət-nə banə 2m an-kəfət-xə banə an-kəfət-xu banə 2f an-kəfət-xʸ banə an-kəfət-xɨma banə 3m an-kəfət-ə banə an-kəfət-o banə 3f an-kəfət-əc banə an-kəfət-əma banə  an-kəfʷəc(-i) banə

Table 72: Negated Past Perfective of √kft ‘open’

160 3.16 Depalatalization Depalatalization stands for the phenomenon that palatalized consonants are replaced by their plain counterpart in some verb forms. All possible consonant pairs palatalized vs. plain (i.e. depalatalized) are listed in table 73.

+ - plosives c t cʼ tʼ j d kʸ k kʼʸ kʼ gʸ g fricatives ʃ s ʒ z xʸ x glide/liquid y r

Table 73: Palatalized vs. plain consonants

Historically speaking, it is probably more reasonable to view this process the other way round. Palatalization emerged diachronically when the above mentioned plain consonants were followed by y/i or w/u (i.e. the abstract phonemes I and U respectively). Accordingly, it is the plain forms that represent the initial state which then became palatalized in most verb forms apart from a few exceptions. Nevertheless, due to the fact that synchronically the forms containing palatalization are in the majority (including the 3sm Perfective which commonly is considered the citation form in Semitic languages including here), the absence thereof is considered here as the ‘exception’. For this reason and in accordance with Banksira (2000: 56ff.) the designation depalatalization has been chosen. Two distinct cases of depalatalization in the verbal paradigms have to be distinguished. On the one hand, there are type B verbs, which all have a root of the shape √123, and on the other hand (most) verbs that have a weak final radical I or U, i.e. √12I, √12U, √123I and √123U. As described on page 52, all type B verbs throughout contain a palatal element which is represented by the superscript  in the root √123. It palatalizes either the first or the second radical or, if this is not possible, the first vowel is raised from ə to e. In the Jussive base, however, the palatalized elements lose this feature completely, i.e. they occur depalatalized. In (463) three example verbs (3sm) illustrate the contrast between palatalized Perfective and Imperfective as against the depalatalized Jussive.

161 (463)    √dβr jəppərəm yɨjəppɨr yədəppɨr ‘finish’ √rkr nəkkʸərəm yɨrəkkʸɨr yərəkkɨr ‘win a lawsuit’ √mtʼr mettʼərəm yɨmettʼɨr yəməttʼɨr ‘select, clean grain’ Note that depalatalization of type B verbs affects the Jussive base, i.e. it occurs in all persons. Compare the full Imperfective and Jussive conjugation of the (weak) verb cot √tUt ‘work’ in table 74. For a formal explanation of depalatalization in the Jussive refer to Banksira (2000: 57f.).

  1s ə-cot nɨ-tot 2sm tɨ-cot tot 2sf tɨ-coc toc 3sm yɨ-cot yə-tot 3sf tɨ-cot tɨ-tot 1p nɨ-cot-nə nɨ-tot-nə 2pm tɨ-cot-o tot-o 2pf tɨ-cot-əma tot-əma 3pm yɨ-cot-o yə-tot-o 3pf yɨ-cot-əma yə-tot-əma  yɨ-coc-i yə-toc-i

Table 74:  and depalatalized  of √tUt ‘work’

Derived stems do not always depalatalize, such as t-stems and at-stems with the infixed reciprocal marker -a- (see table 41) and most of the Frequentatives except the Frequentative t-stems (see table 49). It is important to note, however, that there is some variation as to the occurrence or absence of depalatalization especially in the derived stems, but also in the basic type B verbs. I have recorded several instances of retention of palatalization in the Jussive as for example yəβetɨt ~yəβətɨt ‘let it be wide’. It is not clear whether these are regular (though less frequent) variants, influences from other Gurage dialects, or mere occasional performance mistakes. It is also thinkable that infrequent verbs rarely occur in the Jussive, which may be the cause of confusion (the same is true for mutation patterns). Further, note that there are verbs with palatalized consonants that do not belong to type B and therefore do not depalatalize, for example type A ʃaβ √ʃAβ ‘pull’, type C cʼappər √cʼβr ‘pare root of əssət’ where the palatals are basic root consonants. A different distribution of depalatalization show weak verbs, both triradicals and quadriradicals, with a final radical I or U. As outlined above and illustrated again (with 3sm) in (464), they palatalize the preceding penultimate radical (if palatalizable).

162 (464)    √sxI səkkʸəm yɨsəxʸ yəsxʸ ‘flee’ √fsU fʷəʃʃəm yɨfʷəʃ yəfʷʃ ‘fart’ √kʼmtʼI kʼɨməccʼəm yɨkʼməcʼ yəkʼəmcʼ ‘be ashamed’ √kʼrtʼU kʼʷɨrəccʼəm yɨkʼʷrəcʼ yəkʼʷəncʼ ‘take a handful’ All verbs featuring this format depalatalize in the plural forms with the subject suffixes -(əβ)o¹⁴² and -əma, i.e.  3pm/3pf as well as  and  2pm/2pf and 3pm/3pf; further in the Impersonal (↗ 3.11.5) and the Infinitive with -ot (↗ 3.13). Consider the full paradigm of the bəkkʸə √βxI and the plural forms exhibiting depalatalized k/x instead of kʸ/xʸ in table 75.

   1s bəkkʸə-xʷ-ɨm ə-βəxʸ nɨ-βxʸ 2sm bəkkʸə-xə-m tɨ-βəxʸ bɨxʸ 2sf bəkkʸə-xʸ-ɨm tɨ-βəxʸ bɨxʸ 3sm bəkkʸə-m yɨ-βəxʸ yə-βxʸ 3sf bəkkʸə-c-ɨm tɨ-βəxʸ tɨ-βxʸ 1p bəkkʸə-nə-m nɨ-βəxʸ-nə nɨ-βxʸ-ɨnə 2pm bəkkʸə-xu-m tɨ-βəx-o bɨx-o 2pf bəkkʸə-xma-m tɨ-βəx-əma bɨx-əma 3pm bəkkə-βo-m yɨ-βəx-o yə-βx-o 3pf bəkkə-ma-m yɨ-βəx-əma yə-βx-əma  bəkkəw-i-m yɨ-βəxʷ-i yə-βxʷ-i

Table 75: √βxI ‘cry’ and depalatalization

There are some scattered instances lacking depalatalization which seem to occur rather randomly. Nonetheless, even though they are not the rule, forms like aʃʃəmam ‘they () saw’ (in addition to assəmam ~assəβəmam, but seemingly not *aʃʃəβəmam) do not seem impossible.¹⁴³ Apart from these occasional exceptions, some verbs with final radical I or U regularly never depalatalize. This concerns verbs with penultimate radicals that are not or cannot be palatalized at all as for example the type A verbs kʼʸəmmə √kʼmI ‘win’ or seffə √sfI ‘sew’. They both feature non-palatalizable labials as penultimate radical and instead show palatalization of the preceding consonant kʼ→kʼʸ and vowel raising ə→e respectively. In the depalatalizing contexts this kind of verb does not depalatalize as the 3pm

¹⁴²Verbs with final I or U most frequently have -βo in 3pm of the Perfective, but -o (deleting the base final ə) is also attested, for example cənom ~ cənəβom ‘they came’. ¹⁴³For instance, I have recorded nəccʼəβom / yɨrəcʼo / yəncʼo ‘pluck’ without depalatalization instead of expected nəttʼəβom / yɨrətʼo / yəntʼo. This verb exists also in Amharic (nəccʼə), but since it is found in many Gurage varieties it does not seem to be a loan (cf. Leslau 1979c: 449, Banksira 2000: 227). I believe that the absence of depalatalization here was triggered by the Amharic forms (which stay palatalized throughout) and that normally depalatalization takes place.

163 forms in (465) show. The reason for this must be explained by the fact that the palatal elements are not adjacent to the subject suffixes -(əβ)o and -əma.

(465) √kʼmI ‘win’ √sfI ‘sew’  kʼʸəmmom (*kʼəmmom) seffom (*səffom)  yɨkʼʸəmo (*yɨkʼəmo) yɨsefo (*yɨsəfo)  yəkʼʸəmo (*yɨkʼəmo) yəsifo (*yəsfo)

Further, note that verbs with final radical U and a penultimate radical that is not palatalizable do not surface with a palatal element at all (for example tʼəppʷə √tʼβU ‘suck’ or sɨxʷənə √sxrU ‘be thorny’). When the penultimate radical is palatalizable as well as labializable (i.e. a velar, but only kʼ is attested), there is always labialization (for example nəkkʼʷə √rkʼU ‘shout’ or tʼɨrəkkʼʷə √tʼrkʼU ‘be(come) dea’) and never palatalization (*nəkkʸə, *tʼɨrəkkʼʸə). Evidently, even though belonging to the same type of verbs with final I or U, depalatalization cannot apply in these cases. Verbs of type A with root √1rI, exhibiting the penultimate radical r, represent a special case. While r itself palatalizes to y, its mutated form n does not change. Thus in the Perfective, which features mutation, depalatalization does not apply. In the relevant plural forms of the Imperfective and Jussive, on the other hand, the depalatalized r surfaces, illustrated with xənə √xrI ‘dig a hole’ in (466), as opposed to the palatalized y (which fuses with the preceding vowel, i.e. Imperfective xərI → xəy → xe and Jussive xɨrI → xɨy → xi).

(466) √xrI ‘dig a hole’ 3sm 3pm  xənəm xənəβom  yɨxe yɨxəro  yəxi yəxro

In a similar way, verbs with the root √1rAI (↗ 3.5.3.2) like sɨyə ‘buy’ depalatalize, too, revealing the second radical y→r (467). Note the appearance of vowel a in the Perfective.

(467) √srAI ‘buy’ 3sm 3pm  sɨyəm sɨraβom  yɨsyə yɨsrəβo  yəsəyə yəsərəβo

The monoradicals ʃə ‘want’ and cə ‘leave’ are also depalatalized in the relevant plural forms. Remarkably, however, they also depalatalize in the whole Jussive base as can be seen in (468). Thus they combine the characteristics of verbs with final radical I or U and type B verbs.

164 (468) ʃə ‘want’ cə ‘leave’ 3sm 3pm 3sm 3pm  ʃəm saβom cəm taβom  yɨʃə yɨsəβo yɨc yɨtəβo  yəsay yəsaβo yətay yətaβo It is not clear to what root these verbs belong (if it is possible and useful to determine a root at all),¹⁴⁴ but concluding from above insights one could propose √sIAI and √tIAI. However, this does not explain the difference between ʃə and c of the Imperfective bases. According to Banksira (2000: 227), the Imperative 2sm of cə ‘leave’ in Chaha is tɔ with rounding of the vowel (i.e. tɔ < taw), a fact which suggests a final radical U. Since I have also heard to alongside the recorded tay, it remains to check if one or the other form is more common, and/or if to in Gumer is actually (a contraction o) the Amharic təw(əw) ‘leave (it)ǃ’ that exists as a borrowed exclamation/interjection besides the actual Imperative tay. As mentioned above, the Impersonal uses the depalatalized base, but note that since palatalization is part of the formation of the Impersonal (↗ 3.11.5), verbs with final radical I or U and penultimate alveolars appear again with palatalization in the Imperfective and Jussive as illustrated with the verb accʼə √atʼI ‘shut’ (469). Compare with the verb akkʼʸə √akʼI ‘crunch grain’ (470) which does not feature palatalization of the penultimate velar in the Impersonal.

(469) √atʼI ‘shut’ 3sm 3pm   accʼə-m attʼə-βo-m attʼəw-i-m  y-acʼ y-atʼ-(əβ)o y-acʼ-i  yə-cʼ yə-tʼ-(əβ)o yə-cʼ-i

(470) √akʼI ‘crunch grain’ 3sm 3pm   akkʼʸə-m akkʼə-βo-m akkʼəw-i-m  y-akʼʸ y-akʼ-(əβ)o y-akʼʷ-i  yə-kʼe¹⁴⁵ yə-kʼ-əβo yə-kʼəw-i¹⁴⁶

¹⁴⁴Leslau (1979c: 570) discusses possible origins and cognates of ʃə and tends to relate it to the Amharic type B verb səyyə ‘desire, crave, lust after’. Since this is a type B verb with a root (in Amharic) √syy or √syw, this proposition seems reasonable and the Gumer root √sIAI (or maybe √sIAU) not far fetched (but of course this does not explain the assumed A). ¹⁴⁵Here, the palatalization is ‘absorbed’ by the vowel ə→e (cf. verb type 12I and table 16). ¹⁴⁶According to my recordings, the shorter Jussive 3pm yəkʼo and Impersonal yəkʼʷi seem to be clearly dispreferred.

165 3.17 The verb bar ‘say’ The verb bar √βAr ‘say’ is probably the most frequent verb (apart from the copula) and occurs in a variety of different constructions and functions. The versatility of  is one of the characteristics of the Ethiopian language area (cf. Cohen et al. 2002). Meyer (2009) is an overview of the quotative verb in Ethiosemitic languages and Oromo. It contains data from the Gurage variety Muher and provided some input for this section. Apart from its use as regular full verb, bar ‘say’ primarily functions as quotative verb that marks the end of quoted (direct) speech, either sentences or larger passages. As a further development, the converb of bar grammaticalized (or is on its way to grammaticalize) into a complementizer of verbs of saying, thinking, etc. in general as well as some other subordinating functions, including the formation of clauses to express beneficiaries (see example (491)) and adverbial phrases. In addition, bar is used abundantly in so-called phrasal verbs as the (semantically rather empty) device to integrate non-verbal elements like particles, ideophones, etc. into the verbal system. Finally, bar expresses some ideosyncratic meanings. Formally, probably due to its frequency, bar is slightly irregular in that the Perfective base loses the r before subject suffixes beginning with x (471). Also, it features the unusual vowel ə in the Jussive base (472) (also ↗ 3.5.3.6).

(471) bar-ə-m vs. ba-xə-m say.PFV-3smS-M say.PFV-3smS-M ‘he said’ ‘you (sm) said’ (472) yə-βər vs. yə-xar 3smS-say.JUS 3smS-know.JUS ‘let him say’ ‘let him know’

3.17.1 bar ‘say’ as quotative verb In principle there is no reported speech in Gumer in the sense that the deictic center of a quoted sentence is fully adjusted to the new situation. Rather, the usual way of quoting someone’s speech is by reproducing it as it (presumably) was uttered originally. The quotation is then often (but not obligatorily) closed by the appropriate form of bar ‘say’ marking it as quotation. Consider the following examples (473)-(476).

(473) a-n-ar bar-ə-no-m. NEG-1sS-go.IPFV say.PFV-3smS-3pmO-M ‘He said to them: “I do/will not go.”’ (474) nocʼ sevenap nem bar-o-ya-m. run.IMP.2sfS S. bring.IMP.3sfS say.PFV-3pmS-3sfO-M ‘They said to her: “Run! Bring Seven-up!”’

166 (475) mɨʃt-xʸɨta səβlə bɨ-t-βɨn-na aβet bar-əc-ɨm. woman-DEF.sf S. TEMP-3sfS-say.IPFV-3sfO at.your.service say.PFV-3sfS-M ‘When the woman said to her “Seble”, she said “At your service!”. (476) mʷena-na bə-sost zənga tɨ-ʃr-o-n; maternal.uncle-1sPOSS INSTR-three thing 2S-be.pleasing.IPFV-pmS-1sO ə-rəmdɨ-xu ba-xʷ-ɨn-ɨm. 1sS-love.IPFV-2pmO say.PFV-1sS-3smO-M ‘I said to him: “My uncle (i.e. referring to the related Gyeto people), I like you by (because o) three things; I love you [all]”.’

Gumer being a verb-final language, the quotative verb bar always follows the (whole) quotation, which can be one or several sentences as in the examples above. Similarily, the Impersonal of bar – Perfective bʷarim ‘one told’ or Imperfective yɨwri ‘one tells’ – often closes a longer passage of a narration, marking it as a non- witnessed, retold story. The following excerpt (477) of a narration shows twice an Impersonal of bar after a longer stretch of speech, framed by { and }. As seen in both instances, it can happen that two different forms of bar follow each other, the first one being the quotation verb belonging to the last sentence(s) and the final Impersonal one marking the narration as hearsay.

(477) a. { xɨ cənəw-i-m-tanə bərr ankʷakkʷə-m DEM come.PFV.IPS-3smO-CV.M-LINK gateѦ knock.PFV.IPS-CV.M gəppʷə-m oj-i-m. enter.PFV.IPS-CV.M tell.PFV.IPS-3smO-M ‘{ Then they came and knocked on the door, entered and spoke.’ mɨr gən-l-o? b-i-βɨr gʷəmarə-l-o what country-COP-3pmS TEMP-3smS-say.IPFV Gumer-COP-3pmS bʷar-i-m. say.PFV.IPS-3smO-M ‘When he said: “What country are they [from]?”, they (i.e. servants) said: “They are Gumer”.’ məttaya tʼəβətʼ-o-m cənə-βo-m. bribe take.PFV-3pmS-CV.M come.PFV-3pmS-M ‘“They came with a bribe.”’ mhm. mɨkʼar-u? ɨnde əz-o bar-ə-m. mhm what-COP.3smS please see.IMP-2pmS say.PFV-3smS-M ‘He said: “Mhm. What is it? Please [go and] seeǃ”’ […] zɨkka məgəra. like.that calves ‘“Like this calves.”’ mɨraxɨl-l-o? how.many-COP-3pmS ‘“How many are they?”’

167 arβa yɨ-xər-o. forty 3S-become.IPFV-pmS ‘“Around forty.”’ ɨngʷəd kʼar enə-no-we? other THING NEG.EX[.3smS]-EX-3pmO-Q ‘“Don’t they have anything else?”’ xʷet wəfer nər-ə-no. two young.bull EX-3smS-3pmO ‘“They have two young bulls.”’ tənəf-o-m; afər-əxʷna gəppa-m, gʷəmarə, remain.PFV-3pmS-M land-3pmPOSS enter.PFV[.3smS]-M Gumer bar-ə-m } yɨ-wr-i. say.PFV-3smS-M 3S-say.IPFV.IPS-3smO ‘He said: “They win (remain/escape unharmed); their land entered [the new boundaries], the Gumer” }, it is said.’ b. […] { xɨx məgəra b-osəd-ə-βo ankʼʸə gʷəmarə-xɨno DEM calves LOC/TEMP-take.PFV-3smS-MAL.3pm after Gumer-DEF.pm attɨm səβ t-e-aʒ-o bə-x mədər əj-əxʷna any person TEMP-NEG.3S-see.IPFV-pmS LOC-DEM place hand-3pmPOSS y-amʷəkʼ-o-m y-ɨyya-y t-e-cən-o-m 3S-warm.JUS-pmS-M DAT-1s-GOAL TEMP-NEG.3S-come.IPFV-pmS-ALSO kʼawa-xno yə-tkʼaw-o-m-tanə bet-əxʷna coffee-3pmPOSS 3S-drink.JUS-pmS-CV.M-LINK house-3pmPOSS yə-kr-o bar-ə-m } bʷar-i-m. 3S-ascend.JUS-pmS say.PFV-3smS-M say.PFV.IPS-3smO-M ‘{ Then, after he had taken the calves from them, he said: “The Gumer, before anybody sees them, they should warm their hands over there and without even coming to me they should drink their coffee and go home” }, it was said.’

3.17.2 bar ‘say’ as complementizer / subordinator The verb bar ‘say’, usually in form of a converb, functions as a complementizer of verbs of saying, thinking, etc. and subordinator for various other subordinated clauses. The minimal structure, as it is exemplied in (478), can be summarized as [“” saying ], i.e. it consists of a quoted sentence concluded by a converbal form of bar followed by the matrix verb.

(478) [ “” bar.  ] cənə-xʷ-ɨm ba-xʷ-ɨm od-xʷ-ɨn-ɨm. come.PFV-1sS-M say.PFV-1sS-CV.M tell.PFV-1sS-3smO-M ‘I told him that I came.’

168 The subordinating converb of bar and the matrix verb always share the same subject (3sm in (479)). However, the quoted sentence can have a different subject marking even though referring to the same person, since the deictic center usually is not adjusted to the new speech situation:

(479) cənə-xʷ-ɨm bar-ə-m od-ə-m. come.PFV-1sS-M say.PFV-3smS-CV.M tell.PFV-3smS-M ‘Hei told: “Ii came”.’ → ‘Hei told that hei came.’

The verb bar ‘say’ never combines with the converb of itself in embedding the quotation but stands alone:

(480) cənə-xʷ-ɨm (*ba-xʷ-ɨm) ba-xʷ-ɨn-ɨm. come.PFV-1sS-M say.PFV-1sS-CV.M say.PFV-1sS-3smO-M ‘I said him that I came.’

The verbs of saying, thinking, etc. that occur with the converb of bar as complementizer include od ‘tell’ (481), atcʼawəd ‘chat, talk to, tell stories’ (482), təsar ‘ask’ (483), assəβ¹⁴⁷ ‘think’ (484) and wəssən¹⁴⁸ ‘decide’ (485).

(481) abba-na mʷət-ə-m bar-ə-m y-ud-xə-te. father-1sPOSS die.PFV-3smS-M say.PFV-3smS-CV.M 3smS-tell.IPFV-2smO-FUT.DEF ‘He will tell you that his father died.’ (482) tərakəβ-nə-m bar-ə-m atcʼawəd-e-m. meet.PFV-1pS-M say.PFV-3smS-CV.M tell.stories.PFV-3smS.1sO-M ‘He told me that they met.’ (483) əz-o bar-ə-m təsar-ə-m. see.IMP-2pmS say.PFV-3smS-CV.M ask.PFV-3smS-M ‘He asked [them] to [go and] see.’ (484) əxʷa ɨyya mɨr wə-tot nər-ə-βi ba-xʷ-ɨm assəβ-xʷ-ɨm. now 1s what INF-work EX-3smS-MAL.1s say.PFV-1sS-CV.M think.PFV-1sS-M ‘I thought: “Now, what do I have to do?”’

¹⁴⁷I have recorded assəβ ‘think’ only geminated, also in the Imperfective and Jussive. Since in Amharic assəβ is a type B verb featuring gemination throughout, it seems obvious that it is a loan in Gumer taken over with full gemination (↗ 3.5.3.12). Nevertheless, Leslau (1979b: 620, 1979c: 95) lists asəβ with singleton s in Chaha, which would be the expected format also in Gumer (and in other Gurage varieties). If it is not a loan, the gemination in Gumer could be explained with Amharic influence or, probably more likely, as compensatory lengthening due to the “weakness” of the first (A) and third (β) radicals, comparable to anna √ArA ‘defecate’ or gʸətəβ ~ gʸəttəβ √gtβ ‘bar’ (cf. the discussion on retained gemination on page 47). The alternative verb atcʼamət ‘think, reflect’ (Leslau 1979c: 182) I encountered only once. ¹⁴⁸wəssən ‘decide’ is a loan from Amharic but commonly used. According to Leslau (1979b: 172) the original word for ‘decide’ in Chaha is antʼ (= ‘cut’), metʼəs (= ‘detach’) or tənəm tərəssa (= ‘swearing stand up’?). I have not encountered them in the meaning ‘decide’. It has to be verified if and how frequent they are in use.

169 (485) yə-xno bəz-ɨm tɛ-tə-x-no-tanə fərəz-əna tʼɨβcʼɨ-t-ɨxʷ, DAT-3pm here-ALSO leave-CV.T-1sS-3pmO-LINK horse-1sPOSS take-CV.T-1sS wə-kra nər-ə-βi ba-xʷ-ɨm wəssən-xʷ-ɨm. INF-ascend EX-3smS-MAL.1s say.PFV-1sS-CV.M decideѦ.PFV-1sS-M ‘I decided that I have to go up, leaving them here and taking my horse.’

The following example (486) shows two points to note. First, the overt subject can intervene between the subordinating converb and the matrix verb. Second, as it is the case with quotations discussed above, the verb of saying can be preceded by a longer part of quoted speech consisting of several sentences.

(486) əgi ʃərət ɨm-bəna-xu. əkkʷa mɨʃt tɨ-kʼəttʼɨr-xu-te. okay food PROHIB-eat.PFV-2pmS today woman 3sfS-kill.IPFV-2pmO-FUT.DEF mɨʃt gʷəpsa-n-ya. tɨ-kʼəttʼɨr-xu-te bar-o-m fʷɨga woman stepmother-COP-3sfS 3sfS-kill.IPFV-2pmO-FUT.DEF say.PFV-3pmS-M Fuga od-o-m. tell.PFV-3pmS-M ‘“Ok. Don’t eat the food! Today the woman will kill you. The woman is a stepmother. She will kill you”, the Fugas told.’

3.17.3 Other subordinate clauses with bar ‘say’ The same structure of a quoted sentence subordinated to a matrix verb by means of the converb of bar ‘say’ is also used to express functions other than ‘purely’ quoting speech. Since it is a property of the converbs to be rather vague semantically, the logical connection between subordinated bar-clause and matrix clause tends to be determined from context, world knowledge or verb semantics. For example, a purposive meaning is achieved with a Jussive (1s or 1p) in the “quoted” sentence as in the following examples (487)-(488). For comparison consider the literal translations in brackets. This is only one possiblity to form purposive clauses (↗ 4.7.3.6).

(487) nɨ-ʒ-ɨn bar-ə-m wɨyə-m. 1sS-see.JUS-3smO say.PFV-3smS-CV.M go.down.PFV[.3smS]-M ‘He went down to see him (lit. he went down saying: “Let me see him”).’ (488) nɨ-mʷəkkɨn-ne bar-o-m wəttʼ-o-m. 1pS-try.JUS-1pS.3smO say.PFV-3pmS-CV.M go.out.PFV-3pmS-M ‘They went out to try him (lit.: they went out saying: “Let’s try him”).’

If one changes the TAM form of the “quoted” verb, the interpretation of the logical connection between the clauses changes slightly. While in the first example (489) the Jussive prompts the purposive reading as in the examples above, the Im- perfective in (490) is more neutral and rather interpreted as cause. However, for a more accurate understanding of such differences more data are needed.

170 (489) asa nɨ-tʼɨβtʼ bar-əc-ɨm mərəβ tʼəβətʼ-əc-ɨm. fish 1sS-take.JUS say.PFV-3sfS-CV.M net take.PFV-3sfS-M ‘She took a net to catch fish (lit.: she took a net saying: “Let me catch fish”).’ (490) asa ə-tʼəβtʼ bar-əc-ɨm mərəβ tʼəβətʼ-əc-ɨm. fish 1sS-take.IPFV say.PFV-3sfS-CV.M net take.PFV-3sfS-M ‘Wanting to catch fish she took a net (lit.: she took a net saying: “I catch fish”).’

Beneficiaries can be expressed as complements of bar ‘say’ in the construction ʽ “for ”ʼ, usually with bar as converb as in (491).

(491) y-axə bar-əc-ɨm-ta kʼərikʼar xəna-c-wə. DAT-2sm say.PFV-3sfS-CV.M-LINK little put.PFV-3sfS-MAL.3sm ʽShe put (only) a little bit in it for you (your benefit).ʼ

3.17.4 Phrasal verbs with bar ‘say’ Like all Ethiosemitic and surrounding languages (cf. Meyer 2009), Gumer has a great number of phrasal verbs consisting of a nominal part, often ideophones or other invariable words (sometimes derived from verb roots) plus an auxiliary verb. The most frequent one is bar ‘say’, followed by amənə ‘do’, whereby they often form a pair intransitive vs. transitive, respectively. The auxiliary verbs themselves are semantically rather empty and function as bearer of the verbal information such as TAM and participant marking (cf., for example, Meyer 2006: 100). The list in (492) contains phrasal verbs with bar combined with ideophones and other invariable words that are not relatable to verbs (verb roots) or other nouns. While some of them are clearly of onomatopoethic nature (for example taʔ bar ‘pop’), others are ‘isolated’ instances of (unusual) words, as for example sənkalle bar ‘bind foreleg and hindleg together’. Note that full or partial reduplication or final gemination occur rather often.

(492) Ideophones and invariable words bəkβək bar ‘smell bad (esp. goat, sheep)’ fʷɨnkʸɨnn bar ‘move a little bit aside’ fʷɨtt bar ‘drink at a gulp’ ga bar ‘dawn’ (> gat ‘dawn, early morning’) (ɨn)kʼʷɨss bar ‘be silent, be quiet, keep still’ jad bar ‘struggle’ kəʃə bar ‘have diarrhoe’ (experiencer verb) ko bar ‘scream, shout, shout for help’ (> kot ‘shouting’) kʼo bar ‘stop raining’ kʼərkʼər bar ‘be(come) active, alert’ (experiencer verb) kʼəttʼ bar ‘be straight, upright; stand still’¹⁴⁹ kʼɨr(r) bar ‘feel bad about something, be discontented’ (exper. verb)

171 sənkalle bar ‘bind foreleg and hindleg so that horse (etc.) cannot run’ sɨnkʼʷɨn bar ‘bleed from the nose’ (experiencer verb) ʃɨgɨg bar ‘shudder with disgust’ (experiencer verb) ta~taʔ bar ‘pop’ wəbb bar ‘try, attempt, try out, feign to hit’

In (493) a few examples of phrasal verbs that can be related to verbs or verb roots are shown.¹⁵⁰ Note again the typical reduplicative patterns.

(493) Invariable words attributable to a verb root bʷɨʒbʷɨʒ bar¹⁵¹ ‘feel depressed, feel lonely’ < bazəz ‘feel depressed’ gʸɨrgʸɨr bar ‘blaze, burn easily’ < gʸəkkʸər¹⁵² ‘straighten out’ sɨbbɨrr bar ‘break completely’ < səppər ‘break’ (also Amharic)

Phrasal verbs with əgi ‘okay’ and bay ‘no’ are very frequent. On the one hand, they can mean more or less literally ‘say yes, okay’ (494) and ‘say no’ (495) uttered by a person. On the other hand the subject can also be inanimate (496), in which case əgi and bay bar mean ‘work/function’ and ‘not work/function’ respectively.

(494) əgi bar-əc-ɨm wər-əc-ɨm. okay say.PFV-3sfS-CV.M go.PFV-3sfS-M ‘She said OK and went.’ (495) fʷacʼ bar-e-m-ta əj-əna bɨ-n-artʼ bay mow.IMP[.2smS] say.PFV-3smS.1sO-CV.M-LINK hand-1sPOSS TEMP-1sS-cut.IPFV no ba-xʷ-ɨm. say.PFV-1sS-M ‘He said to me “mow!”, and when I cut my hand I said no (i.e. I stopped, I refused).’ (496) sankʼa bay bar-e-m. door no say.PFV-3smS.1sO-M ‘I could not handle (open/close/etc.) the door. / The door did not work.’

It is even possible to combine inanimate subjects with the structure of quotations. Typically consisting of a negated first person Imperfective, they expresses that something did not work or function (497) similar to bay bar ‘say no’ but thanks to the “quoted” verb specifying the kind of event.

(497) ʃərət-əxʸta a-n-təm bar-ə-m. food-3sfPOSS NEG-1sS-be.tasty.IPFV say.PFV-3smS-M ‘Her food did not become / was not tasty.’ (lit.: ‘Her food said “I am not tasty”’).

¹⁴⁹Cf. Amharic kʼəttʼɨta ‘straight’. ¹⁵⁰Presumably, this list is not representative and could be extended but I lack the necessary data. ¹⁵¹Leslau (1979c: 170) provides buʒʒ bar. ¹⁵²< *√gIrgIr with loss/assimilation of r (↗ 3.5.2.3).

172 The productive construction consisting of bar and a totally reduplicated noun is used express that something smells or tastes like the noun in question. Consider the following illustrative examples:

(498) asso asso bar ‘taste like salt, be salty’ < asso ‘salt’ mʷɨz mʷɨz bar ‘taste/smell like banana’ < mʷɨz ‘banana’ tən tən bar ‘smell like smoke’ < tən ‘smoke’ (499) ɨxa asso asso yɨ-βɨr. water salt salt 3smS-say.IPFV ‘The water tastes salty.’

There are also phrasal verbs with pairs of single (1V22) vs. reduplicated (1V21V2) ideophones, the latter usually expressing intensitiy or repetition. Note, however, the third pair which contrasts with respect to the kind of posture they refer to.

(500) kacc bar ‘move/walk somehow quickly’ kackac bar ‘move/walk very quickly’ təkk bar ‘drip, trickle in drops (once, a bit)’ təktək(k) bar ‘drip, trickle in drops (continuously, a lot)’ zəff bar ‘sit gracelessly (when very exhausted)’ zəfzəf bar ‘walk gracelessly (when very exhausted)’

3.17.5 The infinitives wəβəru and bɨrotu The infinitives of bar ‘say’ plus the enclitic 3sm copula -u, i.e. wəβər-u and bɨrot- u, are used in the meaning of ‘that means, that is to say, namely’ (501)-(502) and as pragmatic marker approximately corresponding to ‘actually’ (503), the boundary between these two functions not being clearcut.

(501) ɨngʷəd yə-dənəs-ə-n-xʷɨta e-ʃɨr-βɨndə other REL-sing.PFV-3smS-3smO-DEF.sm NEG.3smS-be.pleasing.IPFV-MAL.1p kʼar-u a-n-zrəkʸ-ne wə-βər-u. THING-COP.3smS NEG-1pS-tell.IPFV-1pS.3smO INF-say-COP.3smS ‘The other things he sang are not nice, that means we won’t tell them.’ (502) təxankʼʸe dəgmo datta yɨ-frəkʼʷe, zɨ mədər bər-ot-u. afterwards butѦ chest 3S-split.IPFV.IPS.3smO DEM place say-INF-COP.3smS ‘And then the chest is split, that is at this place.’ (503) təxankʼʸə bə-jəppər-xu ankʼʸə mɨkʼar-u afterwards LOC/TEMP-finish.PFV-2pmS after what-COP.3smS yə-tənəf-ə wə-βər-u. REL-remain.PFV-3smS INF-say-COP.3smS ‘And then, after you have finished, what’s then actually?’

173 3.17.6 Further idiosyncratic uses of bar ‘say’ The imperative of bar, used alone, is employed as a request to the conversation partner to start, continue or repeat an action.

(504) bərǃ say.IMP[.2smS] ‘Do it againǃ Go aheadǃ’

Often the intended action denotes specifically beating (cf. Meyer 2009: 32), as it is obvious with the prepositional phrases bəβoks ‘with a box, with the fist’ in (505) and bətifə ‘with a slap in the face’ in (506). (Compare the latter example with example (232), where bətifə is used in combination with the verb ʽhitʼ.)

(505) bə-βoks bʷar-ə-n-ɨm. INSTR-box say.PFV-3smS-3smO-M ‘He punched him with a box (blow with fist).’ (506) bə-tifə bar-əc-na-m. INSTR-slap.in.face say.PFV-3sfS-3sfO-M ‘He slapped her in the face.’

The semantically bleached auxiliary bar is also the default verb used to integrate extralinguistic information and signs. These include body movements like hand gestures or facial expressions, often when demonstrating or imitating an action. In (507), a throwing gesture accompanied the word ɨkka ‘like this’.

(507) naβa-ta tʼəβətʼ-xʷ-ɨn-ɨn-ta cʼəxʷə-na ɨkka ba-xʷ-ɨm. waist-3smPOSS take.PFV-1sS-3smO-CV.M-LINK spear-1sPOSS like.this say.PFV-1sS-M ‘Holding it on its waist (in the middle part) I made like this with my spear.’

3.18 Tense, aspect, and mode (TAM) Gumer distinguishes between three verb bases constituting the three basic TAM forms Perfective, Imperfective and Jussive (including Imperative). The first two can further be furnished with the auxiliary banə ~-βa ‘was, .’ for past tense (↗ 3.18.4), and the latter two with -te and -ʃə, respectively, for two different future tenses (↗ 3.18.5). The Perfective in affirmative main clauses obligatorily features the main verb marker -m, as described in section 3.18.1.1 below.

3.18.1 Perfective The Perfective is used to expresses the perfective aspect, which also covers the functions of the perfect aspect/tense (in contrast to Amharic, for example, there is no distinct perfect verb form, but see example (519) for the experiential perfect). In Bertinetto’s (2006: 266) words “[p]erfectivity refers to an event’s being viewed in its entirety, that is, as a terminated event” entailing “that the speaker envisages the terminal point as an essential part of the event”.

174 From this characterization follows that (total-)terminative (508) and action verbs (509) commonly have a past reading (i.e. when used as an event placed on an actual time-line). Time adverbs like tɨrama ‘yesterday’ in the latter example can additionally underline the past time frame.

(508) təxankʼʸe at ʃi bɨrr aβ-o-m. afterwards one thousand birr give.PFV-3pmS-M. ‘Afterwards they gave 1000 birr.’ (509) tɨrama wɨssa fakʼ-nə-m. yesterday wussa scrape.PFV-1pS-M ‘Yesterday we scraped wussa-bread.’

Since the Perfective refers to a terminated event as a whole irrespective of its internal duration, it is also compatible with verbs and adverbs that express long time spans (510).

(510) səβat sat əkkəs-nə-m. seven hour wait.PFV-1pS-M ‘We waited for seven hours.’

A very large group of verbs in Gumer constitute the inchoative-stative verbs. They are characterized by a change of state – i.e. they have an initial instead of a terminal point – and the following state. The Perfective selecting this initial point expresses that the change of state is terminated, the result of which is the subsequent (new) state. Due to the fact that this state is understood as still lasting (and relevant) at the time of speaking, inchoative-stative verbs usually receive a present reading with the Perfective. For example, kʼʷəm (511) expresses the change of state ‘stand up’ and the resulting state thereof is that someone ‘is standing’ and from cona ‘sit down’ follows that someone ‘is sitting’ (512).

(511) kʼʷəm-o-m. stand.up.PFV-3pmS-M ‘They are standing.’ (512) sost gɨred at-əxʸta afər cona-c-ɨm at-əxʸta b-ombər three girl one-3sfPOSS ground sit.PFV-3sfS-M one-3sfPOSS LOC-chair cona-c-ɨm at-əxʸta kʼʷəm-əc-ɨm. sit.PFV-3sfS-M one-3sfPOSS stand.PFV-3sfS-M ‘Three girls, one (of them) is sitting on the ground, one (of them) is sitting on a chair and one (of them) is standing.’

While ‘stand/stand up’ and ‘sit/sit down’ are undoubtedly belonging together in English forming a pair of a stative and an inchoative-stative verb that is expressed with one lexeme in Gumer, there are many other cases that are not readily recognizable as inchoative-stative verbs (from an English point of view). Consider for example the verb cʼor that is usually translated as ‘carry (on shoulders or head)’, but actually it also includes (or rather primarily means) the (change of state) event

175 preceding the carrying, i.e. the loading (on shoulders or head). Thus, the Perfec- tive in (513) literally can be rendered as ‘(has) loaded’, but translates rather as ‘is carrying’.

(513) nɨkʼyə gɨβɨr cʼor-ə-m. big goods carry.PFV-3smS-M ‘He is carrying a heavy load.’

Other examples include for instance tʼəβətʼ ‘grasp, take → hold’ or the numerous verbs that correspond to adjectives in English, usually as pairs of ‘become V → be V’, such as gənəz ‘be(come) old’.

(514) gɨn əxʷa gənəz-əc-ɨm. butѦ now be.old.PFV-3sfS-M ‘But now she is old.’

Furthermore, probably all experiencer verbs belong to the inchoative-stative verbs. The state, describing bodily sensations, emotions, thoughts, perceptions and the like, the experiencer is in at the actual time of speaking is also rendered by the Perfective, i.e. it is presented as a result. Note again that the Perfective translates as present tense in English.

(515) kʸɨtta kʼəmmʷə-n-ɨm. rabbit fall.ill.PFV[.3smS]-3smO-M ‘The rabbit is sick (more lit.: the rabbit has fallen ill).’ (516) kʼʸətʼ-xʷ-ɨm. be.tired.PFV-1sS-M ‘I am tired (more lit.: I have become tired).’

The Perfective also covers functions of a perfect. Similar to the resulting states of the inchoative-stative verbs, perfects express “the continuing relevance of a previous situation” (Comrie 1976: 56). This reading of the Perfective is thus also possible with terminative and action verbs. Depending on the context, example (517) can refer to an event situated completely in the past, or as perfect with a resulting and still relevant situation. Often the perfect implies a translation with ‘already’ as in (518).

(517) kʼawa sɨyə-xʷ-ɨm. coffee buy.PFV-1sS-M ‘I bought coffee (back then, at that time).’ ~ ‘I have bought coffee (and it is available now).’ (518) kʼənəs-o-m? begin.PFV-3pmS-M ‘Have they (already) begun?’

176 There is a special construction to express an experiential perfect, which “indicates that a given situation has held at least once during some time in the past leading up to the present” (Comrie 1976: 58). It consists of a converb (↗ 3.14) followed by the Imperfective of xar ‘know’ as in (519) (repeated from page 156).

(519) amerika wər-xu-m tɨ-xʸr-o? A. go.PFV-2pmS-CV.M 2S-know.IPFV-pmS ‘Have you (ever) been to America?’

3.18.1.1 Main verb marker -m In affirmative main clauses the Perfective obligatorily features the main verb marker -m. It occurs word-finally after subject and object markers. In contrast, negated and subordinated Perfective lacks -m as illustrated in (520).

(520) main verb subordinated verb  cənə-xʷ-ɨm bə-cənə-xʷ come.PFV-1sS-M COND-come.PFV-1sS  an-cənə-xʷ b-an-cənə-xʷ NEG-come.PFV-1sS COND-NEG-come.PFV-1sS

This morpheme has been analyzed as past tense marker as well (cf. Rose 2007), because it has the same distribution as the two future markers  + -te and  + -ʃə (↗ 3.18.5): they are present obligatorily in affirmative main clauses but absent in negation and in subordination. Here, however, the designation main verb marker is preferred because  + -m never alternates with bare  in main clauses, while bare  and  without the future markers exist (with other meanings of course). Moreover the Futures, in particular the Indefinite Future, are rather modal categories than pure tense. Finally, the past tense reading of the Perfective is not (primarily) triggered by the marker -m but rather inferred from the aspectual (perfective) value. At any rate, here the neutral gloss  is used. The main verb marker should not be confused with the marker of the m- converb (↗ 3.14.2), the focusing -m, or the coordinating -m (ʽʼ).¹⁵³ Consider, for example, the subordinated conditional in (521), where the suffixed -m is a focus marker (ʽalsoʼ), translating as ʽeven ifʼ.

(521) bə-sənəkʼ-xʷ-ɨm neβa e-wɨn-n. COND-steal.PFV-1sS-ALSO thief NEG.3S-say.IPFV.IPS-1sO ʽEven if I steal one will not call me a thief.ʼ

¹⁵³Nevertheless, all -m presumably have the same etymological origin.

177 3.18.2 Imperfective The Imperfective is used to express the imperfective aspect. In contrast to perfec- tivity, “imperfectivity refers to the terminal point of the event not being envis- aged” (Bertinetto 2006: 266), which often is also described as “explicit reference to the internal structure of a situation, viewing a situation from within” (Comrie 1976: 24). In the majority of the cases, the (bare) Imperfective expresses a habitual meaning (522) and has – to a lesser extent – also progressive readings, at least in the sense that an event is going on at the time of speaking (523). There is, however, also an explicit way of forming a high-focal progressive with the copula -u suffixed to the Imperfective, focusing the verb and its imperfectivity (524b).

(522) gʸəta-m gʷəmarə-m yɨ-twakka. Gyeto-ALSO Gumer-ALSO 3smS-fight.IPFV ‘The Gyeto and the Gumer fight with each other (regularly).’ (523) mena-we əxʷa-m yɨ-cot. work-DEF now-ALSO 3smS-work.IPFV ‘He still works /is still working (on the same work).’ (524) a. kʷɨtara yɨ-tʼəβtʼ? chicken 3smS-take.IPFV ʽDoes he catch chickens?ʼ b. kʷɨtara yɨ-tʼəβtʼ-u? chicken 3smS-take.IPFV-COP.3smS ʽIs he catching chickens (right now)?ʼ

Inchoative-stative verbs can only be understood as habitual in combination with the Imperfective (525), since an actual state is expressed by the Perfective.

(525) kʼawa b-an-tɨkʼaw-xʷ gʷɨnər-əna yɨ-fərtʼ-e. coffee COND-NEG-drink.coffee.PFV-1sS head-1sPOSS 3smS-split.IPFV-1sO ‘When I don’t drink coffee, I get a headache.’

Permanent qualities, without any implication that there has been a preceding change of state, can also be rendered by the Imperfective as in (526), apparently interchangeable with the corresponding adjective (plus copula) as in (527).

(526) ɨntʼar-əta y-atʼɨr. stick-3smPOSS 3smS-be.short.IPFV ‘His/the stick is short.’ (527) ɨntʼar-əta accʼɨr-u. stick-3smPOSS short-COP.3smS ‘His/the stick is short.’

Finally note that the Imperfective cannot refer to future events, the choice between one of the two Futures being obligatory (↗ 3.18.5).

178 3.18.3 Jussive The Jussive, to which also the Imperative belongs, is a modal verb form used to express orders, wishes, intentions, permissions and similar meanings. It can only occur in matrix sentences and is never a subordinate mood. Consider the following illustrative examples.

(528) yə-səβat_bet_gʷɨrage ammətʼatʼ-əta n-od-xə. ATTR-Sebat_Bet_Gurage originѦ-3smPOSS 1sS-tell.JUS-2smO ‘Let me tell you the origin of the Sebat Bet Gurage.’ (529) wəfer ərtʼ-oǃ young.bull cut.IMP-2pmS ‘Slaughter a young bullǃ’ (530) danə yə-fɨrd-ɨndə. judge 3smS-judge.JUS-1pO ‘May a judge judge us / A judge shall judge us.’

3.18.4 Analytical forms with past auxiliary banə ~-βa Both the Perfective and the Imperfective (but not the Jussive) form analytical forms with the past auxiliary (‘was’) rendering the event anterior to another one or placing it in the (more remote) past. They can be labelled ‘Past Imperfective’ and ‘Past Perfective’. As shown in (531), there are two interchangeable forms of the auxiliary, the full form banə and a shortened one -βa. The latter seemingly behaves like a clitic and is therefore written adjacent. Both forms are invariable (in contrast to the copula, ↗ 3.19). Note that the main verb marker -m (↗ 3.18.1.1) of the Perfective is not lost.

(531)  wər-ə-m → wər-ə-m banə ~ wər-ə-m-ba ‘he went’ ‘he had gone’  y-ar → y-ar banə ~ y-ar-βa ‘he goes’ ‘he used to go’

The past auxiliary banə ~-βa places the habitual or progressive meaning of the Imperfective into the past, mostly corresponding to English ‘used to’.

(532) mʷeʃə yɨ-cʼəkʼʷs banə. hide 3smS-beg.IPFV AUX.PT ‘He used to beg for hide (i.e. clothes made of dried leather).’ (533) dɨrə ɨkka y-amʷər-i banə? formerly like.this 3smS-do.IPFV.IPS-3smS AUX.PT ‘In the past, did they use to do it like this?’ (534) ɨruz tɨ-cəkɨr-wə-βa. rice 3sfS-cook.IPFV-MAL.3sm-AUX.PT ‘She used to cook rice with/in it.’

179 (535) ə-βɨr-xə-βa. 1sS-say.IPFV-2smO-AUX.PT ‘I used to say [it to] you.’

The past auxiliary with the Perfective results in a form similar to a pluperfect that, broadly speaking, locates a past event anterior to another past event. Accordingly, in (536) the second question of speaker A refers to an event prior to the one in the first question, i.e. emra ‘last year’ as opposed to this year respectively.

(536) A: gən-e an-wənd-ɨxə-we? country-GOAL NEG-go.down.PFV-2smS-Q ‘Did you not go (down) to the countryside?’

B: əə… amət e-xər […] tə-wənd-ɨxʷ? um year NEG.3smS-become.IPFV ABL-go.down.PFV-1sS ‘Um… isn’t it a year since I went (down)?’

A: yə-zər wənd-ɨxə-m-ba emra? DAT-rainy.season go.down.PFV-2smS-M-AUX.PT last.year ‘Did you go (down) for the rainy season last year?’

With inchoative-stative verbs, as it is the case with the experiencer verb in (537), the pluperfect is used to refer to the state when it is not actual any more, the simple Perfective expressing a state lasting at the time of speaking.

(537) balləfə əgr-əta kʼərikʼar akkʼʸə-n-ɨm-ba. latelyѦ leg-3smPOSS a.little.bit hurt.PFV[.3smS]-3smO-M-AUX.PT ‘Lately his leg/foot was a little bit hurt.’

However, it is not always required to backshift a past state with banə as shown in example (538), which even contains the time adverbial tɨrama kʼɨrərə ‘yesterday morning’.

(538) tɨrama kʼɨrərə gʷɨnər-əna fəntʼ-e-m. yesterday morning head-1sPOSS split.PFV-3smS.1sO-M ‘Yesterday morning I had a headache.’

Rather, in such cases the past auxiliary explicitly indicates that the state in question is not lasting any longer as in (539), whereas the lack of it as in (538) above leaves this option open.

(539) gʷɨnər-əna fəntʼ-e-m-ba. təxankʼʸətə akimbet wər-xʷ-ɨm. head-1sPOSS split.PFV-3smS.1sO-M afterwards hospital go.PFV-1sS-M ‘I had a headache. Then I went to the hospital.’

The characterization also applies to verbs other than inchoative-stative verbs. In example (540) the pluperfect wərxʷɨm banə expresses that the speaker went (or had gone) to the town of Awasa but now he is not there anymore.

180 (540) ɨkka tɨ-t-βɨr-e yə-zəgəd-e kʼar awasa like.this TEMP-2smS-say.IPFV-1sO REL-remember.IPFV-1sO THING A. wər-xʷ-ɨm banə, balləfə, t-at samt yɨfte. go.PFV-1sS-M AUX.PT latelyѦ ABL-one week before ‘When you say this to me, what I remember, I went to Awasa, lately, one week ago.’

Finally, banə ~-βa also combines with the Jussive in the apodosis of irreal or counterfactual conditinals (541) (↗ 4.7.3.5).

(541) məkina tə-cənə ʃəwa nɨ-wən-nə-βa. car COND-come.PFV[.3smS] Addis.Ababa 1pS-go.JUS-1pS-AUX.PT ‘If a car came, we would go to Addis Ababa.’~ ‘If a car had come, we would have gone to Addis Ababa.’

3.18.5 Future tense Future events have to be expressed obligatorily by one of the two distinct Fu- ture forms traditionally labeled Definite Future and Indefinite Future (Hetzron 1996, 1977: 85). The Definite Future is formed by -te suffixed to the Imperfective, whereas the marker of the Indefinite Future is -ʃə added to the Jussive base. The subject affixes are the same for both Future forms, namely the ones of the ‘plain’ Imperfective. Hence, as shown in table 76, apart from the addition of -te there is no difference in the conjugation between the Definite Future and the Imperfective (cf. table 57).

  1 ə-kəftɨ-te nɨ-kəftɨ-te 2m tɨ-kəftɨ-te tɨ-kəft-o-te 2f tɨ-kəfcɨ-te tɨ-kəft-əma-te 3m yɨ-kəftɨ-te yɨ-kəft-o-te 3f tɨ-kəftɨ-te yɨ-kəft-əma-te  yɨ-kəfʷc(-i)-te

Table 76: Definite Future (√kft ‘open’)

In contrast, as illustrated in table 77, the conjugation of the Indefinite Future differs from the ‘plain’ Jussive, which features some specific subject markers. In particular, this concerns the 1st person singular ə- vs. n-, the 2nd persons t- vs. Ø- (i.e. the Imperatives), and the 3rd persons y- vs. yə- (cf. table 58).

181   1 ə-kɨftɨ-ʃə nɨ-kɨftɨ-nə-ʃə 2m tɨ-kɨftɨ-ʃə tɨ-kɨft-o-ʃə 2f tɨ-kɨfcɨ-ʃə tɨ-kɨft-əma-ʃə 3m yɨ-kɨftɨ-ʃə yɨ-kɨft-o-ʃə 3f tɨ-kɨftɨ-ʃə yɨ-kɨft-əma-ʃə  yɨ-kɨfʷc(-i)-ʃə

Table 77: Indefinite Future (√kft ‘open’)

The Future markers are suffixed to the very end of a verb form. In case there is an object marker it occurs in its regular place directly after the base (i.e. before the Future marker). Consider for example the 3sm -i of both Impersonals in tables 76 and 77 above. Note that here the object marker is represented in brackets because it is the default marker needed when there is no object (↗ 3.11.5), but in case another object is present -i is “replaced” as illustrated in (542). Consider also example (543) with an object and a verb other than an Impersonal.

(542) y-aw-kə-ʃə. 3S-give.JUS.IPS-2smO-FUT.IDEF ‘One will give [it to] you.’ (543) tɨ-kʼətʼɨr-xu-te. 3sfS-kill.IPFV-2smpO-FUT.DEF ‘She will kill you.’

It goes without saying that the epenthetic vowel ɨ is needed when the suffixes create a succession of three or more consonants (↗ 2.3.2). For instance, with the verb kəfət ‘open’ in the tables above or with sənəkʼ ‘steal’ in (544) ɨ comes between the base final consonant and the initial consonant of the suffix, whereas with səppər ‘break’ (545) its position is before the base final consonant.

(544) /y + sərkʼ + te/ → yɨ-sərkʼɨ-te /y + srkʼ + ʃə/ → yɨ-sɨrkʼɨ-ʃə (545) /y + səβr + te/ → yɨ-səβɨr-te /y + sβr + ʃə/ → yɨ-sβɨr-ʃə

As for the origins of the two markers, it seems to be quite obvious that -ʃə grammaticalized from the verb ʃə ‘want’ (Hetzron 1977: 85). Note, however, that the Jussive does not function as a subordinate mode in dependent clauses, neither with ʃə ‘want’ nor other verbs,¹⁵⁴ i.e. there is no construction {(dependent) Jussive

¹⁵⁴Banksira (2000: 251) uses the term Subjunctive for “what is usually called the Jussive”. In my view the label Subjunctive is highly misleading (if not wrong). It is true that the Jussive in Gumer and the Subjunctives in some European languages (especially Romance) partially overlap in their use

182 + Main Verb ‘want’} that could have served as a direct model for the Indefinite Future. As for the possible origin of -te, the situation is not clear. The final e could be connected to the purposive or directional marker -e (see table 94). Sometimes the pronunciation of -te is rather closed tending to -ti, a fact which is reminiscent of the allomorph -y (< -i) of the purposive or directional marker -e. The origin of the dental t is even less clear. One could try to establish a connection to the dental that optionally occurs in copular forms between personal pronouns and copula (for example ɨyya-t-ɨn-xʷ 1s-t--1s ‘it is me’, see table 79). In Banksira’s (2000: 9f.) synchronic analysis, t is the default “epenthetic” consonant, but in other languages of the Ethiopian language area, t (along with n) occurs rather often as a copula or copula-like element and subsequently also in focus constructions (see most articles in Crass & Meyer 2007, and, for example, Zaugg-Coretti 2009). If this assumption is true, the original meaning of -te could have been something like ‘it is to’, expressing that an event ‘is (about) to’ occur. Anyhow, this scenario is highly speculative and could prove completely wrong by further research.

3.18.5.1 Use of the two Futures

3.18.5.1.1 Future vs. Imperfective In affirmative main clauses the use of either the Definite Future -te or the In- definite Future -ʃə is obligatory for state of affairs that occur after the time of speaking. Consequently the ‘plain’ Imperfective does not cover the whole range of Non-Past but is restricted to what can be called Present tense. This includes events that are going on at the very time of speaking on the one hand (546), and habitual events (547) and generic statements (548) on the other hand.

(546) ete-n-xʸ? — məsəlal acən. where-COP-2sfS ladderѦ [1sS.]bring.IPFV ‘Where are you? — I am getting (or bringing) a ladder.’ (547) zax aβəʃa-xʷɨta ɨnnɨm kərə coc-lə-m yɨ-βəra. DEM habasha-DEF.sm every day work.PFV.IPS-BEN.3sm-CV.M 3smS-eat.IPFV ‘[As for] this habasha, every day one cooks for him and he eats.’ (548) yɨ-gəd-no dengʸa wəfram gɨred yɨ-rəmd. 3smS-be.hungry.IPFV-3pmO boys fatѦ girls 3smS-love.IPFV ‘Skinny (lit. who are hungry) boys love fat girls.’

In contrast, the ‘plain’ Imperfective cannot stand in situations describing future events. Consider for example sentence (549) which contains the temporal expression ‘next year’ explicitly referring to a time point in the future.

(for example as Optative), but the Jussive never appears in dependent clauses, which in turn is the typical occurrence of Subjunctives.

183 (549) yɨ-cən amət yɨ-jəpr-o-te. 3smS-come.IPFV year 3S-finish.IPFV-pmS-FUT.DEF ‘They will finish next year.’

For both Futures there are only affirmative forms and they only occur in main clauses. For negation the negated Imperfective is used (↗ 3.15.2). Thus (550) can stand for habitual and future events.

(550) bə-xʷɨya bɨrr a-n-cot. INSTR-twenty birr NEG-1sS-work.IPFV ‘I will not work for twenty birr.’ ‘I do not work for twenty birr (usually/generally).’

3.18.5.1.2 Definite Future vs. Indefinite Future The difference between the Definite and the Indefinite Future is essentially an epistemic one. With the choice of -te or -ʃə the speaker declares the degree of commitment to the truth or (maybe more adequately) the degree of certainty that something will happen. When a future event is considered certain, the Definite Future is used. The speaker knows that it will definitely happen because it has previously been decided on, is not questionable anymore and/or is inevitable (due to external factors). In other words the future event is presented as a fact. The Indefinite Future, on the other hand, is chosen when a future event cannot be considered certain or given. As stated in Hetzron (1996: 109) it is more subjective than the Definite Future and expresses different nuances of epistemic modality like hopes, intentions, wishes, etc. Often, the essence of the respective meaning can be captured by (implicitly) adding the adverbs ‘certainly’ to the translation of a Definite Future and ‘probably’ or ‘hopefully’ to an Indefinite Future (cf. Banksira 2000: 251, who uses “we know” and “we hope”). Of course this is not equally applicable in all cases, but it is a convenient means to express the basic notion of certainty (551) vs. uncertainty (552).

(551) yɨ-cən-te. 3smS-come.IPFV-FUT.DEF ‘He will come (certainly / definitely / I know / I assure you).’ (552) yɨ-tən-ʃə. 3smS-come.JUS-FUT.INDEF ‘He will come (probably / quite likely / I hope / I assume).’

What follows are several different situations illustrating the use of the two Fu- tures.¹⁵⁵ Typical instances of uncertainty include contexts where the speaker cannot commit to the truth of the state of affairs due to lack of (direct) evidence. The above sentence (552), for example, was uttered to me while we were sitting for

¹⁵⁵For some further examples in several other Gurage varieties refer to Hetzron (1996).

184 some time at a rural road waiting for a bus that could take us from the Gumer countryside to the next town. At that time of the day buses are rare but usually one or the other still shows up. Nevertheless, since noone could have any direct evidence that a bus is indeed on its way (especially in total absence of modern means of communication or official timetables etc.) the only possible choice was the Indefinite Future. If the speaker had had a telephone, for example, and called the driver who then had told them that they are on their way already, the use of the Definite Future as in (551) would have become possible. The two Futures can also distinguish promises (“certain”) from suggestions (“uncertain”). In (553) the speaker indicates with the use of the Definite Future that he is willing to hold his promise. He assures that what he says has been decided already and there is no reason to have any doubts. In contrast, sentence (554) is a suggestion rather than a promise. The speaker had an idea to help out and thus proposes to do the cooking. In addition, since it was the first time for the speaker to cook that specific meal, the Indefinite Future also means that she cannot promise a good outcome but just wants to try to do it.

(553) ʃərət-m-axə dəmoz-m-axə ɨnnɨkʼar-axə ɨyya food-ALSO-2smPOSS salary-ALSO-2smPOSS everything-2sPOSS 1s ə-kəs-te bʷar-ə-nɨ-m. 1sS-pay.IPFV-FUT.DEF say.PFV-3smS-3smO-M ‘He said to him “I will pay for your food and your salary, everything of you”.’ (554) yə-sost səβ ʃərət ɨyya ə-tot-ɨn-ʃə ba-x-na-m. ATTR-three person food 1s 1sS-make.JUS-3smO-FUT.INDEF say.PFV-1sS-3sfO-M ‘I said to her “I will cook the food for three people”.’

In (555) there is an Indefinite Future directly followed by kʼal kɨdanu¹⁵⁶ ‘it is a promise’, a fact which seems to contradict the above description. However, here the promise (‘we will take you’) is depending on a condition (‘tell us!’) which has to be fulfilled first in order for the promise to become a “fact”.

(555) yɨna n-an-nə mədər y-axu-m nɨ-sdɨ-ne-ku-ʃə. 1p 1pS-go.IPFV-1pS place DAT-2pm-ALSO 1pS-take.JUS-1pS-2pmO-FUT.INDEF kʼal_kɨdan-u. a-n-c-ɨne-ku. od-o-ndə. promiseѦ-COP.3smS NEG-1pS-leave.IPFV-1pS-2pmO tell.IMP-2pmS-1pO ‘We will take (also) you to the place we are going. It is a promise. We will not leave you. Tell us!’

Another example for the conditional nature of the Indefinite Future is shown in (556). Here the addressee is told to open the door if someone specific comes. Since there is a condition, the opening of the door cannot be considered an event that is certain to happen, a fact which explains the use of the Indefinite Future.

¹⁵⁶kʼal kidan ‘promise’ is a loan from Amharic.

185 (556) sada tɨ-t-cən sankʼa tɨ-kɨftɨ-ʃə. S. TEMP-3sfS-come.IPFV door 2smS-open.JUS-FUT.INDEF ‘When Sada comes you (should) open the door.’

The Definite Future also occurs in questions. At first sight this seems to be counterintuitive since one would normally believe that the speaker, i.e. the one that asks, cannot but be uncertain about the future events as such. Why else would they have to ask? Consider the following short excerpt of a conversation (557) containing one content and one yes/no question.

(557) A: mʷan-mʷan yɨ-cən-te bə-gən? who-who 3smS-come.IPFV-FUT.DEF LOC-country ‘Who (pl) will come / is coming from the countryside?’ B: gaʃʃe.¹⁵⁷ father ‘Father.’ A: gaʃʃe yɨ-cən-te? father 3smS-come.IPFV-FUT.DEF ‘Father will come / is coming?’ B: ɨnk. yes ‘Yes.’

In the first question, speaker A wants to know who will come to town to attend speaker B’s graduation ceremony. At least two factors render somebody’s future coming uncertain: first, the final exams are still months ahead, so actually it is not even certain that B will eventually graduate; second, A does not know if it has already been decided at all that somebody will come (or if it is feasible). Neverthe- less, the Definite Future is the right choice here because in this question speaker A does not mean to express their own certainty about the event but they want to inquire who is already known (to speaker B) to certainly come to the ceremony. The Indefinite Future would rather be used when one is pondering who in general could be considered a possible candidate to come. As to the first factor, by using the Definite Future speaker A respects speaker B’s capability of graduat- ing successfully. The Indefinite Future could rather be interpreted that speaker A does not take the upcoming graduation for granted. Also in the second question of (557) the Definite Future is not referring to the asker’s certainty. In fact this question is not intended to ask for new contents but it is simply an echo question repeating what has been said just before (even though B’s answer did not contain the verb yɨcənte ‘he will come’, this is what was meant). There are instances where the use of the Definite Future is the most adequate choice for pragmatic reasons even though the speaker cannot be certain

¹⁵⁷gaʃʃe is an Amharic term of respect to older males, very often used to refer to one’s father.

186 as such about the future event. In (558), although the speaker has no control of what will happen and moreover even in combination with the modal expression gʷeta bəβarə ‘God willing’, the Definite Future is employed to present the state of affairs as if it were a fact in order to encourage the addressed person in his plans.

(558) axə dəmmo – gʷeta bə-βar-ə – zədrə tɨ-kʼərs-ɨm-ta 2sm butѦ God COND-say.PFV-3smS this.year 2smS-begin.IPFV-CV.M-LINK yɨ-cən amət tɨ-jəppɨr-te. 3smS-come.IPFV year 2smS-finish.IPFV-FUT.DEF ‘You then – God willing – will begin this year and next year you will finish.’

Similarly, the Definite Future occurs in menaces and threats to underline their significance. In example (559) the speaker is kept prisoner and has no information whether his people, the Gumer, will come to free him (i.e. by exchanging calves as ransom) or not. The Indefinite Future here would sound like a harmless suggestion and deprive the sentence of its menacing nature.

(559) gʷəmarə yɨ-cən-te awa-xə məgəra Gumer 3smS-come.IPFV-FUT.DEF woe-2smPOSS calf y-acən-te. 3smS-bring.IPFV-FUT.DEF ‘The Gumer will come – woe to you! – they will bring calves.’

As we have seen so far, the Indefinite Future is used when a speaker is not certain about a future event as with nəgə ‘tomorrow’ in (560).

(560) nəgə sada sukʼ tɨ-kɨftɨ-ʃə. tomorrow S. shop 3sfS-open.JUS-FUT.INDEF ‘Tomorrow Sada will [probably] open a shop.’

However, it can also occur in contexts that refer to past events as in (561). Here, the person who answers the question about an event that presumably took place tɨrama ‘yesterday’ does not know the correct answer but assumes (or infers) that it happened and expresses this by means of the Indefinite Future.

(561) A: sada tɨrama sukʼ kəfət-əc-ɨm? S. yesterday shop open.PFV-3sfS-M ‘Did Sada yesterday open a/the shop?’ B: tɨ-kɨftɨ-ʃə. 3sfS-open.JUS-FUT.INDEF ‘Probably she has opened [one/it].’

This example suggests that the modal (i.e. epistemic) value of the Indefinite Future is stronger than the temporal component – possibly even up to the extent that it is analyzable as the only inherent category. However named, the vast majority of occurrences of the Indefinite Future still refers to future time points. One general

187 exception to this tendency, though, is the use of the copula or more precisely its corresponding (suppletive) Indefinite Future form yɨxɨrʃə etc. based on the verb xər ‘be, become’. In (562) tɨxɨrʃə does not translate as ‘you will be/become’ but it expresses probability.

(562) zɨkka bə-βa-xʷə-n gi axə yə-mʷət-ə abba-na like.this COND-say.PFV-2smS-3smO time 2sm REL-die.PFV-3smS father-1sPOSS balagara tɨ-xɨr-ʃə. opponent 2smS-become.JUS-FUT.INDEF ‘If you say such things about him, you [probably] are my dead father’s enemy.’

Finally, note that the copula does not need to occur in a Future tense (563).

(563) nəgə jɨmat-u. tomorrow Friday-COP.3smS ‘Tomorrow is Friday.’

3.19 Verbs of being The verbs of being form an irregular system. The present tense copula () is enclitic and does not belong to any verb root. Rather it consists of a copular element (-n-/-l-) plus person markers partly corresponding to the  set. The present tense of the verb of existence () is a distinct verb form nər- (not attributed to any root) with the subject markers of the  set. In negation and subordination as well as TAM forms other than present tense, suppletive verbs are employed. On the one hand, there are suppletive forms going back to the common Ethiosemitic root *√hlw ‘exist’ (Meyer 2011: 1246), all of which employ the  subject markers. They are anə- (subordinated ), enə- (negated ), and banə- (past of  and ). The latter also functions as past tense auxiliary in analytic forms with  and  (↗ 3.18.4), however often only in its invariable (enclitic) form -βa. On the other hand, the verbs xər √xr ‘become’ (negation and/or subordination of ) and nəppər √rβr ‘live’ (negation and/or subordination of banə) are used.

  1 -n-xʷ -ndə 2m -n-xə -n-xu 2f -n-xʸ -n-xɨma 3m -u -l-o 3f -n-ya -l-əma

Table 78:  present tense (affirmative)

The present tense copula () consists of -n- plus subject markers in part corresponding to the ones of the  set. For 3pm/3pf the copular element is

188 -l- rather than -n-. In 1p, -n- is absorbed by the nasal of the subject marker -ndə. The 3sm copula does not feature any consonantal marker, but consists of -u only (-w after vowel as in (568)). The copula is a clitic attaching to the last word of the predication as in (563) above or (564)-(568).

(564) gʷɨrage-n-xu. Gurage-COP-3pmS ʽYou (pm) are Gurage.ʼ (565) gʷəmarə-l-o. Gumer-COP-3pmS ʽThey (pm) are Gumer.ʼ (566) nɨkkʼar wəxe kʼar-u. very good THING-COP.3smS ʽIt is very good.ʼ (567) mətʼaf-xʷɨt bə-satʼin fʷər-u. book-DEF.M LOC-box on-COP.3smS ʽThe book is on the box.ʼ (568) mena-xno-w. work-3pmPOSS-COP.3smS ʽIt is their job.ʼ

An additional copular element -t- can occur optionally when  directly follows the corresponding personal pronoun of the first and second persons. The third persons do not seem to allow for -t-.¹⁵⁸

  1 ɨyya-t-ɨn-xʷ yɨna-t-ɨndə 2m axə-t-ɨn-xə axu-t-ɨn-xu 2f axʸ-t-ɨn-xʸ axma-t-ɨn-xɨma 3m xʷɨt-u xɨno-l-o 3f xʸɨt-ɨn-ya xɨnəma-l-əma

Table 79:  with pronouns and -t-

A typical occurence of these forms is in cleft sentences as in (569).

(569) ɨyya-(tɨ-)n-xʷ yə-sɨyə-xʷ-ɨn. 1s-(t-)COP-1sS REL-buy.PFV-1sS-3smO ʽI bought it (lit. it is me who bought it).ʼ

The present tense of the verb of existence () is nər-, which only exists in this form and is not attributed to any verb root. Note that the final r assimilates to the

¹⁵⁸Reportedly, 3smS xʷɨt-ɨt-u ʽit is himʼ is possible in Ezha.

189 nasal of 1pS. Formally,  is conjugated like a  verb (↗ 3.11.1), but since it expresses present tense it does not occur with the main verb marker -m (↗ 3.18.1.1).

  1 nər-xʷ nən-nə 2m nər-xə nər-xu 2f nər-xʸ nər-xɨma 3m nər-ə nər-(əβ)o 3f nər-əc nər-əma

Table 80:  present tense (affirmative)

Some typical uses of the verb of existence are illustrated in (570)-(573).

(570) rəwda nər-əc. R. EX-3sfS ‘Rauda is here (i.e. around, available).’ (571) xɨkka nər-əǃ there EX-3smS ‘It is there!’ (572) b-ɨnəkʼʷamt yəcʼəmar, gəmʷə, togyə, mənzo, gʸɨβyə yɨ-wr-i LOC-Y. Y. G. T. M. G. 3S-say.IPFV.IPS-3smO səβ nər-ə. person EX-3smS ‘In Yinekwamt (a subgroup of Gumer) there are people (i.e. subgroups of Yinekwamt) called Yechemar, Gemwe, Togye, Menzo, [and] Gyibye.’ (573) yɨ-tcʼawəj-i kʼar təcʼər nər-ə. 3S-chat.IPFV.IPS-3smO THING abundantly EX-3smS ‘There are a lot of things to talk about.’

As mentioned above, some suppletive forms of verbs of being go back to the Ethiosemitic root *√hlw ‘exist’. They all have the shape of and are conjugated like a  verb as table 81 shows. Note that they never feature the main verb marker -m, reasons being that they express present tense and/or are subordinated and/or negated. The basic form is anə, which only exists as subordinated .¹⁵⁹ The negated present tense of  is enə¹⁶⁰ ʽthere is notʼ, which is also used in subordination. Finally, banə functions as past tense of both  and  in main clauses. This verb most probably originates in the subordinated form b- (/) + anə, which also accounts for the absence of -m despite its past tense reference. However, other explanations about its origins have been proposed (cf. Hetzron 1977: 106f.).

¹⁵⁹In contrast, for example, to the cognate allə in Amharic, which is  and  in main clauses. ¹⁶⁰Cf. the Amharic cognate form yəllə-m ʽthere is notʼ.

190 Negated  Subordinated  Past  and  1s enə-xʷ -anə-xʷ banə-xʷ 2sm enə-xə -anə-xə banə-xə 2sf enə-xʸ -anə-xʸ banə-xʸ 3sm enə -anə banə 3sf enə-c -anə-c banə-c 1p enə-nə -anə-nə banə-nə 2pm enə-xu -anə-xu banə-xu 2pf enə-xma -anə-xma banə-xma 3pm enə-βo -anə-βo banə-βo 3pf enə-ma -anə-ma banə-ma

Table 81: Verbs of being originating in *√hlw ‘exist’

The following examples illustrate negated  in main clauses (574)-(576) and in subordination (577). Note that the subordinator yə- is missing as is often the case with relativized negated verbs (↗ 4.7.3.1).

(574) neβa enə. thief NEG.EX[.3smS] ‘There is no thief (there are no thieves).’ (575) bə-βet səβ enə. LOC-house person NEG.EX[.3smS] ‘There is nobody at home.’ (576) cʼəxʷə-na tʼəβətʼ-xʷ-ɨm tɨ-n-cən yɨ-kʼʷəm səβ spear-1sPOSS take.PFV-1sS-CV.M TEMP-1sS-come.IPFV 3smS-stand.IPFV person enə. NEG.EX[.3smS] ‘There is no one who stays put/upright when I come with (lit. holding) my spear.’ (577) afinjə enə-wə kʼar chili NEG.EX[.3smS]-MAL.3sm THING ʽsomething without chili in itʼ

In subordination,  is anə as in the complement clause (578) or the relative clause (579).

(578) ɨnnɨmgi bə-dəwəʃe y-anə-xʷ-xəma tɨ-xʸɨr-we? always LOC-D. REL-EX-1sS-COMP 2smS-know.IPFV-Q ʽDo you know that I am always (available/found) in Deweshe?ʼ (579) əxʷa y-anə-xə-wə məsriyaβet? now REL-EX-2smS-MAL.3sm workplaceѦ ʽThe workplace you are in now?ʼ

191 Both  and  (affirmative) share a common past tense banə (.). Exam- ple (580) shows a past tense copula and (581) a past tense verb of existence.

(580) afər-əta tʼəβɨβ banə. (< tʼəβɨβ-u) land-3smS narrow BE.PT[.3smS] ‘His land was narrow.’ (581) bə-fərənji gən at aβəʃa banə. (< at aβəʃa nərə) LOC-foreigner country one habasha BE.PT[.3smS] ‘There was a habasha in a foreign country.’

The same verb banə is also the past auxiliary in analytic forms with  and  (↗ 3.18.4), however quite often in its invariable (enclitic) form -βa (582). For the past of  and  (.), the full form banə is used more frequently, but it also occurs as -βa (583).

(582) angɨwa yɨ-wəre-βa. whey 3S-eat.IPFV.IPS.3smO-AUX.PT ‘They (one) used to eat whey.’ (583) yədrə səβ xʷɨjɨr y-əram goga-βa-ʃ. former person clothes ATTR-cow skin-BE.PT-PRAG ‘The clothes of people of old was cow leather (you know).’

Present tense  is replaced by the  of xər √xr ʽbecomeʼ in negation (584)- (585) and subordination (586)-(587). Note that this suppletive verb expresses present tense despite its  form.

(584) dəβo-axə an-xər-ə. relative-2smPOSS NEG-become.PFV-3smS ʽHe is not your relative.ʼ (585) bə-gən-əxʷna nɨkkʼar yi-kʷəʃ-i, an-xər-ə-we? LOC-country-3pmPOSS very 3S-pay.IPFV.IPS-3smO NEG-become.PFV-3smS-Q ʽIn their country they pay a lot, isnʼt it (i.e. donʼt they)?ʼ (586) xʸɨt dəngənə yə-xər-əc-xəma ə-xʸɨr. 3sf rich REL-become.PFV-3sfS-COMP 1sS-know.IPFV ʽI know that she is rich.ʼ (587) ərəkʼʷe yə-xər-e far REL-become.PFV-3smS.PURP ʽbecause it is farʼ

As suppletive verb for negated and subordinated banə (i.e. past  and past ) the  of nəppər √rβr ‘live’ is used. Examples (588) and (589) show complement clauses of a past tense copula and a past tense verb of existence, respectively, and (590) illustrates negated banə.

192 (588) xʸɨt mərkama yə-rəppər-əc-xəma ə-xʸɨr. 3sf beautiful REL-live.PFV-3sfS-COMP 1sS-know.IPFV ʽI know that she was beautiful.ʼ (589) at gərəd yə-rəppər-əc-xəma ə-xʸɨr. one girl REL-live.PFV-3sfS-COMP 1sS-know.IPFV ʽI know that there was one girl.ʼ (590) tʼɨrə an-nəppər-ə. expensive NEG-live.PFV-3smS ʽIt was not expensive.ʼ

Furthermore, the  of nəppər √rβr ʽliveʼ instead of  can be used to express permanent states (591). Similarly, in combination with an m-converb it describes a habitual state of affairs (592).

(591) bə-gəbya yɨ-rəβɨr. LOC-market 3smS-live.IPFV ‘It is available on the market.’ (592) ɨxa səccʼə-xʷ-ɨm ə-rəβɨr. water drink.PFV-1sS-CV.M 1sS-live.IPFV ‘I (always/habitually) drink water.’

3.19.1 ʽHaveʼ and ʽmustʼ Verbal possession (ʽhaveʼ) is expressed by the verb of existence (and its suppletive forms) plus the primary object suffixes, and obligation (ʽmustʼ) is expressed by the verb of existence plus malefactive suffixes (see table 61 for the forms). In verbal possession, the primary object refers to the possessor while the possessed is the subject (593)-(594). This is evident in example (595) where the verb agrees with the feminine noun. However, usually the verb of existence appears simply as masculine singular (i.e. the default gender, ↗ 4.2.3), no matter if the possessed is plural (596) or even human (597).

(593) wənəxʷə nər-e. neighbor EX-3smS.1sO ʽI have a neighbor.ʼ (594) bet banə-ndə. house BE.PT[.3smS]-1pO ‘We had a house.’ (595) at gʷaddəɲɲa banə-c-e. one friendѦ BE.PT-3sfS-1sO ‘I had one (a certain) friend ().’ (596) xʷet wəfer nər-ə-no. two young.bull EX-3smS-3pmO ʽThey have two young bulls.ʼ

193 (597) dengʸa nər-ə-n. boys EX-3smS-3smO ʽHe has boys.ʼ

When the possessor appears as an overt noun it stands before the possessed noun (598)-(599). Note that it remains unmarked, even though in other instances dative- like primary objects may be marked with yə- (↗ 4.7.1.2).

(598) ɨntəganə sera enə-n. Endegeñ custom NEG.EX[.3smS]-3smO ‘The Endegeñ do not have customs.’ (599) dada lufa gaz enə-n. D. L. war NEG.EX[.3smS]-3smO ‘Dada Lufa does not have [any history o] war.’

Obligation is expressed literally as ʽthere is  on someoneʼ. The content verb is in the form of an infinitive (or verbal noun, ↗ 3.13) which functions as the syntactic subject. The obligated person is indexed as malefactive on the verb of existence (600)-(602).

(600) wə-βra nər-ə-βxə. INF-eat EX-3smS-MAL.2sm ʽYou (sm) must eat (lit. eating is on you).ʼ (601) wə-kra nər-ə-β-i. INF-ascend EX-3smS-MAL.1sO ʽI have to go up.ʼ (602) ətʼβ-ot nər-ə-wə. wash-INF EX-3smS-3smO ʽHe must wash.ʼ or ‘It is necessary to wash ( ~ one must wash).’

Negated  +  yields the translation ʽshould not, need notʼ (603). This example also illustrates that the obligated person (axə) remains unmarked as overt noun or pronoun despite the fact that it is indexed as malefactive on the verb.

(603) xʷɨt bə-mʷət-ə gɨzyə axə wə-mbər enə-βxə. 3sm LOC/TEMP-die.PFV-3smS time 2sm INF-live NEG.EX[.3smS]-MAL.2sm ‘When he is [already] dead, you should/need not live.’

194 4 Nominal morphology

4.1 Derivational nominal morphology In Gumer we can distinguish between two categories of derivational nominal morphology. First, there is a small set of affixes with a more or less consistent usage to derive nouns (and some adjectives) from adjectives and nouns. The most productive ones with a clear meaning are -nət (604), -ina (605), and -ənə (606). Some other affixes can be assigned a meaning, too, but they occur less frequently and are – if at all – less productive (607)-(615). Second, there is a great variety of irregular patterns to derive nouns and adjectives from verbs (or rather verb roots). They are usually neither productive nor attributable to specific meanings.¹⁶¹ Among these unpredictable patterns there are ‘bare’ forms only consisting of the root consonants (622) and forms with additional suffixes as for example -a (616), -ət (618) or -t (620). Note that some of these patterns feature additional palatalization and/or labialization that is not part of the basic root, for example (624) or (625). See Banksira (2000: 185ff.), in particular for his analysis of such root-internal labial- izations and palatalizations,¹⁶² and Rose (2007: 423ff.) for a general short overview of the derivational nominal morphology. The suffix -nət (604) derives abstract nouns from adjectives (a-d) and nouns (e-h). Final vowels are deleted (d-h) and the gemination in kʼʸamma (g) is reduced.

(604) -nət a. mʷəkʼ ‘warm’ mʷəkʼnət ‘heat, temperature’ b. bətɨt ‘wide, broad’ bətɨtnət ‘width, breadth’ c. nɨkʼ ‘big’ nɨkʼnət ‘greatness’ d. dəngənə ‘rich’ dəngənnət ‘richness, being rich’ e. wənəxʷə ‘neighbor’ wənəxʷnət ‘neighborhood’ f. tɨkə ‘child’ tɨknət ‘childhood’ g. kʼʸamma ‘guarantor’ kʼʸamnət ‘guarantee, insurance’ h. amakʸə ‘(wife’s male) amakʸnət ‘(wife’s male) in-law(s)’ kinship’

¹⁶¹Due to the fact that these patterns are rather unproductive, it seems at times difficult to consider them as nominal derivations as such, especially the rare ones. In some cases it might therefore be safer to state that certain nominals and verbs can be identified with the same root. Consider for instance zər ‘seed’ and zəna ‘sow’ (√zrA) or kʼɨβ ‘butter’ and kʼəppa ‘smear’ (√kʼβA). ¹⁶²In short, labialization and palatalization are both triggered by the features [round] and [high], respectively, of an abstract element /U/ word-internally or as (part o) a suffix. I do not intend to reject this analysis, but I will not go into more details here for mainly for two reasons: first, a majority of Banksira’s examples, mainly the so-called participles, seemed to be rather marginal or even unknown to my Gumer speakers and could therefore not be confirmed; second, Banksira’s analysis does not predict all forms correctly (see, for example, Banksira 2000: 204). Thus, in addition to the fact that the derivations in question can hardly be assigned specific meanings, there are many irregularities and exceptions. In order to obtain a completer picture of the nominal derivations in Gumer, a more systematic survey is needed.

195 There are also words with -nət that are not necessarily abstract derivations from a concrete adjective or noun. For instance, we can find the words ʃɨtnət that seems to be derived from ʃɨtta. However, these are coexisting words with apparently the exact same meaning ‘smell, odor’, the latter probably being a loan from Amharic. Another example is the pair tʼor and tʼornət, both seemingly expressing the same or almost the same meaning ‘war, fighting’. Alongside -nət there is also the variant - ɨnnət in use as, for example, tʼorɨnnət ‘war’ and kʼʸammɨnnət ‘guarantee’. Probably, this geminated form is either borrowed directly from Amharic or an influence from a neighboring geminating Gurage variety. The suffix -ina (605) is added to the name of peoples, nations, places or coun- tries to form the corresponding language name. It can also be used with mɨr ‘what’ to ask for a language (d). The final vowel of the basic word is deleted (a,b). Besides -ina, the Amharic form -ɨɲɲa is also in use, especially among younger speakers.

(605) -ina a. gʷɨrage ‘Gurage (people)’ gʷɨragina ‘Gurage language’¹⁶³ b. amara ‘Amhara (people)’ amarina ‘Amharic language’ c. jərmən ‘German, Germany’ jərmənina ‘German language’ d. mɨr ‘what?’ mɨrina ‘what/which language?’

The suffix -ənə (606) added to nominals forms agentive nouns or adjectives, mostly expressing a profession or a (permanent) quality. Note that the final vowel is deleted (d). In the case of ʃɨrənə ‘crazy’ (g), the word is not directly derived from ʃɨr, but is connected to the reduplicated form ʃɨrʃɨr.¹⁶⁴ Finally, the suffix -ənə is also used to form ordinal numerals (h) (↗ 4.5.2).

(606) -ənə a. fərəz ‘horse’ fərəzənə ‘rider’ b. dəm ‘blood’ dəmənə ‘murderer’ c. gaz ‘war’ gazənə ‘warrior’ d. mena ‘work’ menənə ‘worker’ e. naxʷcər ‘message’ naxʷcərənə ‘messenger’ f. tʼom ‘fast’ tʼomənə ‘Christian’ g. ʃɨrʃɨr ‘gotten crazy’ ʃɨrənə ‘mad, crazy’ h. sost ‘three’ sostənə ‘third’

Another suffix occuring in agentive nouns or adjectives is -amma (607). It is much less common than -ənə and probably not productive. Occasionally, it can also be heard without gemination, especially with mərkama ‘beautiful’.

¹⁶³Or gʷɨragʷe / gʷɨragʷina. ¹⁶⁴The possible “candidates” ʃɨr ‘lawsuit’ and ʃɨra ‘blossom of əssət’ have to be excluded due to their semantics. Rather, ʃɨrənə is connected to the reduplicated ʃɨrʃɨr ‘gotten crazy’ used in ʃɨrʃɨr bar ‘be crazy’ (cf. Banksira 2000: 142 and Leslau 1979c: 584).

196 (607) -amma a. ojə ‘gossip, news, lie’ ojamma ‘liar’ b. weg ‘chant, song, poem’ wegamma ‘minstrel’ c. mərk ‘appearance, shape’ mərkamma ‘beautiful, pretty’

It is important to note that by far not all agentive nouns or professions can be derived by -ənə or -amma. More commonly these concepts are expressed by relative clauses attributed to səβ ‘person’, i.e. ‘somebody who is V-ing’ as in (608) and (609) (cf. also example (614)).¹⁶⁵

(608) xʷɨjɨr yɨ-sef səβ clothes 3smS-sew.IPFV person ‘tailor (lit.: a person who sews clothes)’ (609) awi yɨ-tʼəβtʼ-o səβ wild.animals 3S-take.IPFV-pmS person ‘hunters (lit.: persons who catch wild animals)’

The suffix -wət (610) is used to derive feminine nouns (a,b), or nouns with a special related meaning (c). Again, note that final vowels are deleted and gemination is reduced.

(610) -wət a. mərkamma ‘beautiful’ mərkamwət ‘beautiful one ()’ b. gəmbəna ‘dark’ gəmbənwət ‘dark one ()’ c. tʼɨrar ‘shade’ tʼɨrarwət ‘shadow’

According to Rose (2007: 424), the suffix -yə (611) converts adjectives into nouns with an extended meaning as illustrated in (a).¹⁶⁶ However, as for ɨrsɨyə ‘small’ and nɨkʼyə ‘big’ (b,c) it is not clear if there is a difference in meaning. With nouns, -yə occurs together with the prefixed attributivizer yə- (d).¹⁶⁷

(611) -yə a. tʼɨkʼʷɨr ‘black’ tʼɨkʼʷɨryə ‘black wɨssa-bread’¹⁶⁸ b. ɨrs ‘small’ ɨrsɨyə ‘small’ c. nɨkʼ ‘big’ nɨkʼyə ‘big’ d. ɨmmat ‘only one’ y-ɨmmatyə ‘an only child’

¹⁶⁵Not uncommon is also the use of an Amharic loan word, for example asa atʼmaj ‘fisherman’ instead of “original” Gumer asa y-atʼəmd səβ. ¹⁶⁶I could not find more equivalent examples. ¹⁶⁷Rose (2007: 423) mentions that also -wət if used with a noun occurs with yə-. However, I could not find relevant examples (Rose’s example yə-ʃəxɨr-wət ‘potter ()’ was not accepted). ¹⁶⁸As it seems tʼɨkʼʷɨryə can denote any object that is black, but in particular it is used for a black kind of wɨssa.

197 With eβər- ‘so-and-so’ -yə stands for masculine, whereas -wət (see above) marks the feminine form, i.e. eβəryə ‘so-and-so (m)’ and eβərwət ‘so-and-so ()’. The two prefixes wə- (612) and mə- (613) occur in some instrumental nouns. They can contain other affixes as -a, -yə or -t. There are also derivations that convey non-instrumental meanings like mənkəs ‘stomach-ache’.

(612) wə- a. seffə √sfI ‘sew’ wəsifə ‘awl’ b. fəccʼə √ftʼI ‘grind’ wəfcʼə ‘lower millstone’ c. fənt √frt ‘cut in hal’ wəfənca ‘entrance (dividing the house in two)’ (613) mə- a. atrassa √rsA ‘help to lift’ matraʃyə ‘stretcher for carrying dung’ b. kʼʷənə √kʼrU ‘roast’ makʼʷret¹⁶⁹ ‘stick used to roast coffee beans’ c. nəkəs √rks ‘bite’ mənkəs ‘stomach-ache’

It is important to note that such instrumental derivations cannot be applied to any verb root. In fact there are only a handful of pertinent examples (cf. also Banksira 2000: 216). Usually, instrumental meanings are expressed by impersonal constructions in a relative clause (ʽrelative verbʼ, ↗ 4.7.3.1) involving an instrumental object (glossed , ↗ 3.12.1) attributed to the dummy head kʼar ‘’.¹⁷⁰

(614) yɨ-kəfʷcɨ-pʷə kʼar 3smS-open.IPFV.IPS-MAL.3sm THING ‘opener (lit.: a thing with which one opens)’

In addition to all the above affixes that are productive or at least attested several times, there are some derivational suffixes in abstract nouns, like -nat, -nər, -cər, -ərə, -arə (615), that occur very rarely, most of them presumably only once. Note that they are not suffixed to some other noun or adjective but form an abstract noun “directly” from the verb root. For instance there is no such noun as *k’ʸəm or *kʼʸəmmə from which kʼʸəmnat ‘sickness’ would be derived. Rather it is connected to the verb k’ʸəmmə ‘be ill’ and the verb root √kʼmI.

¹⁶⁹The verb makʼʷret should be derived from akʼʷənə with causative a-, but it is not clear if this is in use. ¹⁷⁰Other heads for the ʽrelative verbʼ are also possible. Leslau (1979c: 528) has for example (Chaha) yɨ-rəʃ-pʷə gɨβɨr ‘weaving loom (lit. an object/utensil with which one weaves)’. I have also recorded instances without head noun such as yɨ-gəwə-pʷə ‘entrance (lit. one enters by/in it)’, though in such cases it is not entirely clear if they are really used as referential nominal like this or if they are predications.

198 (615) other suffixes a. kʼʸəmmə √kʼmI ‘be ill’ kʼʸəmnat ‘sickness’ b. kʼʸətʼ √kʼItʼ ‘be tired’ kʼʸɨcʼnər ‘fatigue’ c. nax √rAx ‘send’ naxʷcər ‘message’ d. tʼəkkʼʷər √tʼkʼʷr ‘be black’ cʼəkʼʷrərə ‘early morning’ e. məʃʃə √msI ‘become evening’ mɨsarə ‘evening’

There is a considerable number of nominalizations with a suffix -t, -ət, -at or -a.¹⁷¹ Generally, they are all identifiable with verb roots, but the derivations do not follow specific patterns.¹⁷² The following lists are not exhaustive but show some illustrative examples. Words of similar shape are grouped together. Nominals with the suffix -a (616) often have the pattern 123-a (a-e) and show labialization of the rightmost labial or velar consonant (kʼura in (a) is derived from *kʼɨwra < *kʼɨβʷra), and final alveolars are palatalized (c-d). The example zənga (), derived from the weak quadriradical verb zɨrəkkʸə √zrgI, behaves differently.

(616) -a a. kʼəppər √kʼβr ‘be incomplete’ kʼura ‘incomplete’ b. gəffər √gfr ‘let go, release’ gɨfʷra ‘abandoned (house)’ c. məna √mrA ‘be full’ mʷɨra ‘full’ d. agəz √Agz ‘help’ əgʷʒa ‘ally, help’ e. wənd √Urd ‘go down’ wɨrja ‘abortion’ f. zɨrəkkʸə √zrgI ‘speak’ zənga ‘word, matter’

Another example that seems to fit into this class is gʷɨdβa ‘trench’. Etymologically it is without doubt connected to the verb gʸəttəβ (√gdβ) ‘bar’, but in this case it is most probably a loan from Amharic (gudba) rather than a derivation directly from the Gumer verb or verb root. If it were originally Gumer one could expect the final labial β to be labialized. In contrast to -a, nominals with -at do not feature labialization (617),¹⁷³ but the medial radical r in (a,b) appears in its palatalized form y. Note that except for xəttərat (c) all examples have weak final radicals. Since this is mostly A, the a might also be seen as part of the base and not of the suffix -at. Nevertheless, I assume that here the suffix is indeed -at which deletes the root-final vowels (in accord with Banksira 2000: 196).

¹⁷¹According to Rose (2007: 424) they go back to former gender markers. ¹⁷²As mentioned above, there is an analysis of the derivations involving labialization and palatalization in Banksira (2000: 185ff.); see footnote (162). ¹⁷³The labialized fʷ in (a) is due to the final radical U.

199 (617) -at a. a-fʷənə √frU ‘rest’ fʷəyat ‘rest’ b. a-xəna √xrA ‘shout, make noise’ xəyat ‘noise, shouting’ c. xəttər √xdr ‘stop from fighting’ xəttərat ‘truce’¹⁷⁴ d. gəppa √gβA ‘enter’ gəppat ‘evening’ e. kʼɨyə √krAI ‘wait; look afterʼ kʼərat ‘night watch’ f. tʼəmma √tʼmA ‘be thirsty’ tʼɨmat ‘thirst’ g. wəzza √UzA ‘sweat, be shiny’ wɨzat ‘sweat’

Nominal derivations with -ət (618) often follow the pattern 123-ət without additional labialization or palatalization (a-d), but note the slightly different output with the weak verbs in (e-).

(618) -ət a. fəntʼ √frtʼ ‘have headache’ fɨrtʼət ‘headache’ b. nəppər √rβr ‘live’ nɨβrət ‘life’ c. sədəd √sdd ‘drive away’ sɨddət ‘exile’ d. fəttʼər √ftʼr ‘create’ fɨtʼrət ‘creature, nature’ e. cʼar √cʼAr ‘load’ cʼərət ‘load, freight’ f. cʼɨyə √crAI ‘stink’ cʼinət ‘bad smell’ g. effə √IfI ‘cover with lid’ ɨffʷət ‘pot lid’ h. a-xʷə -¹⁷⁵ ‘leak’ axʷət ‘leak (roo)’

Finally consider the following few examples with final -it or -et (619) and the nouns with final -t that are derived from the ideophones (620) used in phrasal verbs with bar ‘say’ (↗ 3.17.4).¹⁷⁶

(619) -it, -et a. kʼʷənnər √kʼrr ‘trim (tip of house)’ kʼʷɨrit ‘tip of house’ b. tʼənəkʼ √tʼrk’ ‘be dry’ tʼɨnkʼʸit ‘fright’¹⁷⁷ c. seffə √sfI ‘sew’ sɨfet ‘sewing’

¹⁷⁴xəttərat is also the name of a Gumer village. Further, concerning its meaning, Leslau (1979c: 370) states that the equivalent word in Inor, Endegeñ and Gyeto means ‘postponement asked by the family of the killer concerning the payment of blood money’. ¹⁷⁵It is not evident what the rood of xʷə should be. Comparing it to the cognate verb in Geʼez kaʕawa √kʕw (Leslau 1987: 272) it could be √xAU. Then again the y in Jussive base xay in Gumer suggest the presence of a radical I, as do the (Perfective) forms hoʔyə in Endegeñ or kəʕe in Tigre (cf. Leslau 1979c: 362). ¹⁷⁶There are also nouns ending in -it that are not derived from verb roots, for example ɨmbʷɨrβit ‘whirlwind’ or ɨnkʼʷɨrfit ‘obstacle’. ¹⁷⁷tʼənəkʼ ‘be frightened’ is an experiencer verb. The basic meaning is ‘be dry’.

200 (620) -t a. ga bar ‘dawn’ gat ‘dawn’ b. ko bar ‘shout, scream’ kot ‘shouting’

Some nouns feature the shape 12a3-ə with their final root consonant palatalized (621). If it is not palatalizable the preceding vowel a is raised to e (e). As can be seen in (b,c,) the medial radical occurs in its mutated form. When there are weak radicals, the pattern is not evident anymore on the surface, for example gajə < *gAajə (d).¹⁷⁸ Note that there can also be additional labialization (c). It has to be left open if this is due to the fact that the basic verb is a type B verb.¹⁷⁹

(621) 12a3-ə +  a. nəməd √rmd ‘love’ nɨmajə ‘love’ b. dənəg √drg ‘hit’ dɨnagʸə ‘cough’¹⁸⁰ c. ʃəpət √sβt ‘choose’ ʃɨpʷacə ‘choice’ d. gad √gAd ‘be hungry’ gajə ‘hunger’ e. sənəf √srf ‘fear’ sɨnefə ‘fear’

A note on the productivity of such derivations is in order here. Overall, it seems that other than the few suffixes above (604)-(606) the derivational patterns are not productive at all but each of them contains only a limited number of mem- bers. Nevertheless, it is not ruled out that one could form creatively new words based on one or the other pattern. This happened with the pattern 12a3-ə and the verb cəffər √tfr ‘take a mouthful’. In Gumer, there is no noun derived from this root, but to express ‘mouthful’ the Amharic loan gurʃa~ gʷɨrʃa is used (which relates to the Amharic verb gʷərrəs ‘take a mouthful’). Asked if there really is no Gurage equivalent of gʷɨrʃa, a speaker suggested that it had to be cɨffʷarə immediately adding, however, that this word does not exist. Quite a big number of nouns and adjectives derived from verb roots have the shape 123 without additional affixes (622). Many of them, however, feature additional internal labialization (623), palatalization (624) or both (625). When there are weak radicals involved, the connection between form and pattern are not always evident,¹⁸¹ as for instance tʼu and eβ. Note also that a root-final A changes to ə when palatalization or labialization is involved (there is no labializable consonant in tʼɨrə and arə). The following lists are not exhaustive.

¹⁷⁸Another word that might belong to this group is ojə ‘gossip, news’ from od ‘tell’ and the root √AUd, i.e. ojə < *AUajə. On the other hand, why would then the output not be əwajə? See also Leslau (1979c: 111) who has əwjə and not *əwajə in the related Gurage language Kistane (Soddo). ¹⁷⁹This hypothesis is supported by the Chaha word k’ʸɨnawə ‘proximity, nearness’ (Banksira 2000: 218, Leslau 1979c: 491) which fits into this pattern. It is derived from a type B verb (k’ənəβ, √krβ ‘be near’) and also features labialization (β→w). However, it is not sure if the derived noun is also common in Gumer. ¹⁸⁰‘Coughing’ is an experiencer verb, for example yɨ-dərg-e ‘I am coughing’ is literally ‘it hits me’. ¹⁸¹The classification of such cases might be disputed.

201 (622) 123 a. fənt √frt ‘cut in hal’ fɨnt ‘hal’ b. fənd √frd ‘judge’ fɨrd ‘judgment’ c. məkkər √mgr ‘suppurate’ mɨgɨr ‘pus’ d. tʼəkkʼʷər √tʼkʼʷr ‘get black, burn’ tʼɨkʼʷɨr ‘black’ e. xənəm √xrm ‘spend a year’ xɨrɨm ‘year’ f. tʼəppʷə √tʼβU ‘suck’ tʼu¹⁸² ‘breast’

(623) 123 +  a. bəssər √βsr ‘be ripe, cook’ bʷɨsɨr ‘ripe’ b. dəməd √dmd ‘gather, join, unite’ dɨmʷd ‘joined, united’ c. fəkkʼər √fkʼr ‘be fat’ fʷɨkʼʷɨr ‘fat (n)’ d. fəntʼ √frtʼ ‘become blind’ fʷɨrtʼ ‘blind’ e. gədəd √gdd ‘pierce’ gʷɨdɨd ‘torn’ f. məntʼ √mrtʼ ‘peel’ mʷɨrtʼ ‘barren’¹⁸³ g. mezzər √mzr ‘count’ mʷɨzɨr ‘number’ h. nəfəg √rfg ‘be greedy’ nɨfʷɨg ‘greedy’ i. səkkʼər √skʼr ‘hang’ sɨkʼʷɨr ‘roo’ j. tʼəbəs √tʼβs ‘roast’ tʼus¹⁸⁴ ‘roasted (meat)’ k. tʼəffa √tʼfA ‘be spoilt, extinguish’ tʼɨfʷə ‘bad, evil’ l. təppa √tβA ‘be hard’ tɨwə ‘hard’ m. tʼəna √tʼrA ‘be expensive’ tʼɨrə ‘dear, expensive’ n. anna √ArA ‘defecate’ arə ‘stool’

(624) 123 +  a. attʼər √Atʼr ‘be short’ accʼɨr ‘short’ b. tə-fəkkʼʸər √fkʼr ‘sing’ fɨkʼʸɨr ‘song’ c. fəna √frA ‘have intercourse’ fɨyə ‘vagina’ d. təffa √tfA ‘slap; spit’ tifə ‘slap (in face)’ e. anəβ √Arβ ‘milk’ eβ ‘milk’¹⁸⁵

(625) 123 +  +  a. tə-xəttər √xdr ‘dress’ xʷɨjɨr ‘clothes’ b. fətʼəm √ftʼm ‘block up, close’ fʷɨcʼɨm ‘closed, untouched’ c. bettər √βtr ‘distinguish, separate’ bʷɨcɨr ‘different’ d. mettʼər √mtʼr ‘select, clean (grain)’ mʷɨcʼɨr ‘neat, washed (clothes)’

¹⁸²tʼu < *tʼɨw < *tʼɨβʷ (the labialization is due to the final radical U) ¹⁸³Usually understood as a piece of land without grass (because it was eaten by cows etc.). ¹⁸⁴tʼus < *tʼɨws < *tʼɨβʷs ¹⁸⁵eβ < *ayβ

202 Another smaller group of nouns and adjectives has the pattern 1ə2ə3.

(626) 1ə2ə3 a. tʼənəkʼ √tʼrkʼ ‘be dry’ tʼərəkʼ ‘dry’ b. wəttər √Utr ‘draw tight’ wətər ‘tendon’ c. kʼɨməccʼə √kʼmtʼI ‘be ashamed’ kʼəmətʼ ‘shame’ d. gədər¹⁸⁶ ‘new’

There are numerous other patterns for nouns and adjectives, many of them considerably less frequent than the forms presented above. Consider the following two non-exhaustive lists of illustrative examples, divided into forms without (627) and with additional suffixes (628).

(627) a. zənəβ √zrβ ‘rain’ zɨraβ ‘rain’ b. kʼəmmər √kʼmr ‘kill louse’ kʼɨmar ‘louse’ c. nəfəs √rfs ‘blow (wind)’ ɨmfas ‘wind’ d. betət √btt ‘be wide’ bətɨt ‘wide, broad’ e. tʼəβəβ √tʼββ ‘be narrow’ tʼəβɨβ ‘narrow’ f. cʼor √tʼUr ‘carry’ tʼor ‘bundle (of grass)’ g. nakʼ √rAkʼ ‘be bigger’ nɨkʼ ‘big’ h. cʼənə √tʼrI ‘give birth’ cʼɨn ‘woman-in-childbed’ i. gənəf √grf ‘be long’ gef ‘long’ j. mʷakʼ √mʷAkʼ‘be warm’ mʷəkʼ ‘warm’ k. fənəx √frx ‘be patient’ faraxʸ ‘patient’ l. cʼəkʼʷəs √tʼkʼʷs ‘beg’ cʼakʼʷaʃ ‘beggar’ m. cot √tUt ‘work; farm’ cəwac ‘farmer’ n. sassa √sAsA ‘be thin’ sɨsə ‘thin’ o. fəkkʼa √fkʼA ‘split’ fəkʼʷə¹⁸⁷ ‘split’ p. dɨrəttər √drdr ‘thicken’ jənjɨr ‘thick’ q. dɨrəzəz √drzz ‘be blunt’ dɨrzɨz ‘blunt’ (628) a. betət √btt ‘be wide’ bʷɨtɨto ‘worn out cloth’ b. kətəf √ktf ‘chop, hash’ kɨtfʷə ‘chopped raw meat’ c. gəttər √gdr ‘put to sleep’ gɨdyə ‘sleep’ d. bəkkʸə √bxI ‘cry’ bɨxʸə ‘mourning, funeral’ e. a-kʸəs √kIs ‘joke’ kʸəʃə ‘joke’ f. gədəd √gdd ‘pierce’ gʷəjə ‘hole’ g. xənə √xrI ‘dig a hole’ xəyə ‘ditch’

¹⁸⁶The adjective gədər ‘new’ seems the belong to the root √gdr. However, I find it difficult to attribute it to the verb gəttər ‘put to sleep’ based on the same root. Another word that features the same consonants is gadɨr ‘cattle pen inside house’. Since this is the place where cows sleep, the meaning of the verb seems to fit in this case. According to Leslau (1979c: 265), however, this word is a loan from a Cushitic language. Furthermore, I am not aware of any other nominal derived from a verb root featuring the same template 1a23. ¹⁸⁷This is the only example with this shape (first vowel ə with additional ) I am aware of.

203 h. tənə √trI ‘swear’ təyə ‘oath’ i. kʼʷənə √kʼrU ‘roast’ kʼʷərə ‘roasted grain’ j. wənd √Urd ‘go down’ wərəjə ‘forecourt’¹⁸⁸ k. xar √xAr ‘know’ xari ‘wise, intelligent’ l. kʼʷəmmər √kʼmr ‘be strong’ kʼʷəmarə ‘strong, fully grown’ m. cot √tUt ‘work; farm’ cuca ‘ploughing, working’ n. ʒəʃʃə √zIzI ‘be cold’ ziza ‘cold, wet’ o. tʼɨrəkkʼʷə √tʼrkʼU ‘be dea’ tʼənkʼʷalla¹⁸⁹ ‘dea’

4.2 Number and gender Nominal morphology marking gender and number is almost inexistent in Gumer. While the verbal conjugation and pronouns (including the possessive suffixes) distinguish between masculine and feminine gender as well as singular and plural number, nouns show these features overtly only in some limited cases, mainly with the definite articles (which themselves are derived from pronouns, ↗ 4.4.1). In other words, the categories number and gender that are inherent to nouns man- ifest almost only in the agreement with “pronominals” including overt pronouns, possessive suffixes, definite articles, and in particular the verbal subject and object agreement affixes.

4.2.1 Number Nouns do not form any plurals at all except for a very small group that have suppletive plurals (see table 82). The same form is used for singular (629)-(630) and plural reference (631), as well as in generic contexts (632).

(629) bəlay yɨ-wr-i gʷəppay-əna yə-cʼɨʃt-e bar-ə-m B. 3S-say.IPFV.IPS-3smO brother-1sPOSS DAT-Ch.-GOAL say.PFV-3smS-CV.M kəna-m. ascend.PFV[.3smS]-M ‘My brother called Belay said “To Chisht”¹⁹⁰ and went up.’ (630) fərəz-əna tʼɨβcʼɨ-t-ɨxʷ wə-kra nər-ə-βi ba-xʷ-ɨm horse-1sPOSS take-CV.T-1sS INF-ascend EX-3smS-MAL.1sO say.PFV-1sS-CV.M wəssən-xʷ-ɨm. decideѦ.PFV-1sS-M ‘I decided that I have to take my horse and go up.’

¹⁸⁸wərəjə is the space between the house(s) and the fence separating the private property from jəffʷərə (the public space/road in villages). ¹⁸⁹According to Leslau (1979c: 624), tʼənkʼʷalla with geminate ll is only Ezha, whereas the Chaha form of ‘dea’ is tʼənkʼʷara. Since usually Gumer is expected to be more like Chaha than Ezha, it has to be checked which form is more common among Gumer speakers. ¹⁹⁰The cult of cʼɨʃt is the “yearly ceremonial in honour of the Sky-god wakʼ [Waq]” and is attended by men only (Shack 1966: 180).

204 (631) bə-βet, tə-gʷəppay-əna sost fərəz tʼəβətʼ-nə-m wəttʼa-nə-m. LOC-house COM-brother-1sPOSS three horse take.PFV-1pS-CV.M go.up.PFV-1pS-M ‘At home, [me] with my brothers took three horses und went out (set of).’ (632) ɨyya fərəz nəkm-ot yɨ-ʃr-e. 1s horse ride-INF 3smS-be.pleasing.IPFV-1sO ‘I like to ride horses.’

In spite of the fact that nouns are invariable, they have to agree in number (and gender, see below) with the verb. In the following examples the subject (633) and object (634) – clearly recognizable as plurals thanks to the plural definite article - xɨno – reflect the plural number in the subject and object suffixes respectively (and of course the singular subject in (634) agrees with the singular subject marker).

(633) bet-xɨno nɨkkʼal-l-o. house-DEF.pm many-COP-3pmS ‘The houses are many.’ (634) astəmari-xʷɨta yə-təmari-xɨno təsar-ə-no-m. teacher-DEF.sm DAT-student-DEF.pm ask.PFV-3smS-3pmO-M ‘The teacher asked the students.’

To be sure, singular or plural number does not have to be visibly expressed in the nominal morphology (i.e. by the definite article). Rather, number is semantically determined and manifests in the verbal agreement where required. In both (635) and (636) the object is bet ‘house’, but in the former example it is singular and in the latter plural as it can be seen in the corresponding object suffixes on the verb.

(635) y-astəmari-xʷɨta bet aʃʃə-xʷ-ɨn-ɨm. ATTR-teacher-DEF.sm house see.PFV-1sS-3smO-M ‘I saw the house of the teacher.’ (636) y-astəmari-xʷɨta bet aʃʃə-x-no-m. ATTR-teacher-DEF.sm house see.PFV-1sS-3pmO-M ‘I saw the houses of the teacher.’

As mentioned above, a very small group of nouns have suppletive plural forms all belonging to the core vocabulary denoting female and male humans as well as cattle. Banksira (2000: 242) names six cases (for Chaha) as listed in table 82.

¹⁹¹According to Leslau (1979c: 213) dengʸa ‘boys’ might be borrowed from a Cushitic language where the original dungʷi meant ‘servant’. ¹⁹²According to Leslau (1979c: 82), əray ‘cows’ is despite its similarity to the singular əram ‘cow’ a loan from a Cushitic language where it means ‘cattle, domestic animals’. Thus, it might be possible that also in Gumer it is a collective noun rather than a plural in the strict sense.

205   mɨʃt ɨʃta ‘woman/women’ mɨs gəmya ‘man/men’ gərəd gɨred ‘girl/girls’ ərc dengʸa¹⁹¹ ‘boy/boys’ əram əray¹⁹² ‘cow/cows’ (dək ??məgəra ‘calf/calves’)

Table 82: Suppletive plural forms (according to Banksira 2000: 242)

According to my observations, however, only five of them are obligatory number oppositions. The case of məgəra ‘calf/??calves’ is highly questionable and needs more investigations.¹⁹³ Leslau (1979c: 394) translates it as plural ‘calves’, but there are some contradicting facts. Firstly, the alleged singular dək apparently denotes a kind of calf different to məgəra and even Leslau (1979c: 204) says that dək means ‘calf of a certain size’. Secondly, məgəra is used in contexts that are typical of singular. For instance, in generic usage as in (637) and (638) the alleged plural word məgəra is used rather than dək. Along these lines, compare its use in example (638) with sentence (639) where the singular word əram ‘cow’ and not the plural əray ‘cows’ appears in the basically same function of specifying the material.

(637) yədrə məttaya dəmmo məgəra-w-ʃ. former bribe butѦ calves-COP.3smS-PRAG ‘In old times the bribe was calves, you know.’ (638) xa-ta yə-məgəra dannəra banə. DEM-3smPOSS ATTR-calves tanned.hide BE.PT ‘The other ones were calf leather.’ (639) yədrə səβ xʷɨjɨr y-əram goga-βa-ʃ. former person clothes ATTR-cow skin-BE.PT-PRAG ‘The clothes of people of old were cow leather, you know.’

In addition to the above singular usage, məgəra occurs also in unmistakably plural contexts after numbers higher than one as with bəkʼɨr ‘hundred’ in (640).

(640) mʷena-ta togyə bəkʼɨr məgəra yə-sɛ, ɨngʷəd ɨnəkʼʷamt bəkʼɨr uncle-3smPOSS T. hundred calves 3smS-find.JUS other I. hundred məgəra yə-say. calves 3smS-find.JUS ‘Yes, his uncles, the Togye, should find one hundred calves, the other [people of the] Inekwamt should find one hundred calves.’

¹⁹³Menuta (2002), for example, lists the same words in Ezha (with slightly different pronounciation) except ‘cow/cows’. Thus also according to him dəkk and məgəra are suppletive singular/plural forms of ‘calf.’

206 Thus, in conclusion, məgəra means ‘cal’ and is used unchanged for all numbers very much like any other ordinary nouns. Likewise dək, rather than being the singular of məgəra, is a common noun, too, referring to a different type of animal ‘calf of a certain size’. Nevertheless, regardless of the actual status of məgəra, one might postulate that the suppletive plural words do not express number in a strict sense but are, for example, collectives and/or analoguosly the singular forms singulatives. This hypothesis would be supported in the case of məgəra which is etymologically related to məgarya in Tigrinya meaning ‘flock, herd’ (Leslau 1979c: 394) and thus does not refer to single individuals. Be this as it may, the suppletive plurals (except məgəra) always trigger plural (and gender) agreement on the verb. Thus, dengʸa ‘boys’ in (641) agrees with the masculine plural object and subject suffixes -no and -o respectively and gɨred ‘girls’ agrees with -əma in (642), very much as it is the case with all plural nouns like in (633) or (634) above.

(641) yɨ-gəd-no dengʸa wəfram ɨʃta yɨ-rəmd-o. 3smS-be.hungry.IPFV-3pmO boys fatѦ women 3S-love.IPFV-pmS ‘Hungry (slender) boys love fat women.’ (642) arβət gɨred b-ombər con-əma-m tənkʼʸəkʼʸəf-əma-m. four girls LOC-chair sit.PFV-3pfS-CV.M embrace.each.other.PFV-3pfS-M ‘Four girls embraced each other sitting on a chair.’

4.2.2 Associative marker nə- The associative marker nə- (n- before vowels) is a prefix that attaches to nouns (or noun phrases). It is used to express “plurality” in the sense that one refers the noun in question plus similar items or persons that are typically associated with it. Thus, nə-dɨstʼ in (643) referts to pots and other items of the semantic field of dishes like plates, cutlery, drinking glasses etc. without specifying them. In (644), nə- occurs with agaz, a military rank, referring to all men with the same title (in this case nə- is similar to a plural marker) and/or men with the title agaz and others possessing comparable military ranks. The combination of nə- with a proper name like nə-βətrə in (645) translates as ‘Betre and everybody associated with Betre’, which, depending on the context, can mean Betre and his family and relatives, or for example Betre and his followers or group, etc.

(643) nə-dɨstʼ aʃkʼar t-aʃkʼar tɨ-rəkʸ. ASS-pot something COM-something 3sfS-throw.IPFV ‘She throws pots and dishes and things and stuff.’ (644) n-agaz and enə-no. ASS-agaz compulsion NEG.EX[.3smS]-3pmO ‘The agaz and so on do not have a problem. (645) zɨ nə-βətrə metyə-xʷna-w. DEM ASS-B. grandfather-3pmPOSS-COP.3smS ‘This is the grandfather of Betre and the like.’

207 In example (646) the speaker repeated the almost same sentence only changing the object, which nicely illustrates the contrast between an associative noun and a list of single items, i.e. ‘bananas and similar things’ vs. ‘oranges and bananasʼ.

(646) nə-mʷɨz yɨ-srəβ-o-e-w-ɨʃ yə-cənə-βo, ASS-banana 3S-buy.IPFV-pmS-PURP-COP.3smS-PRAG REL-come.PFV-3pmS bɨrtɨkʷan-ɨm mʷɨz-ɨm yɨ-srəβ-o-e-w-ɨʃ orange-ALSO banana-ALSO 3S-buy.IPFV-pmS-PURP-COP.3smS-PRAG yə-cənə-βo. REL-come.PFV-3pmS ‘They came to buy bananas and the like, they came to buy oranges and bananas.’

Examples (645) above and (647)-(648) below show that nə- (like bə- and tə-) deletes the prefix yə-. The suffix -xəma ‘like’ always occurs together with yə- (↗ 4.7.2) but with nə- it misses (i.e. *nə-yə-farda-xəma).

(647) xɨkka nə-farda-xəma soresa bə-tmanəx-ə […] like.this ASS-F.-like hero COND-be.captured.PFV-3smS ‘If like this a hero like Farda is captured…’ (648) nə-βərdəfərə nə-βərkəfətə-xəma-nyə bə-xər-o […] ASS-berdefere ASS-berkefete-like-DIR COND-become.PFV-3pmS ‘If they are something like berdefere [or] berkefete…’

Example (649) shows that nə- also occurs with demonstratives, the attributivizer yə- again deleted (i.e. < yə-z yə-dada ɨndaʃo metyə ‘grandfather of this dada In- dasho’). Note that the double marking of demonstrative and noun is common.

(649) nə-z nə-dada ɨndaʃo metyə-ta-w, mənzo-w. ASS-DEM ASS-dada I. grandfather-3smPOSS-COP.3smS M.-COP.3smS ‘He is a grandfather of this dada Indasho and the like, he is a Menzo.’

Finally, consider nə- with the numeral arβət ‘four’ in (650) where it does not mean ‘four and other numbers’ but ‘four or a similar number’, i.e. ‘about, around’.

(650) mɨraxɨr jəffʷərə wəsəd-ə-βo-m wə-βər-u? wəʃərmɨne, how.much public.space take.PFV-3smS-MAL.3pm INF-say-COP.3smS W. dɨlfay, əxʷa-m səndəkʼʷə, n-arβət. D. now-ALSO S. ASS-four ‘How many jəffʷərə (i.e. villages) did he take from them, then? Wesher- mine, Dilfay, and also Sendekwe, about four.’

4.2.3 Gender Similar to number, also gender is not expressed by nominal morphology (except for very marginal instances, cf. ↗ 4.2.3.1). The only cases where masculine and feminine genders have to be distinguished are again the overt pronouns, posses-

208 sive suffixes, the definite articles and verbal agreement. Gumer features masculine as the default gender. This means that everything is treated grammatically as masculine except for human beings of female sex. Consider the feminine subject and object markers as well as the feminine definite article in example (651) containing two female participants, as well as the feminine plural forms agreeing with gɨred ‘girls’ in example (652).

(651) mɨʃt-xʸɨta səβlə bɨ-t-βɨn-na aβet bar-əc-ɨm. woman-DEF.sf S. TEMP-3sfS-say.IPFV-3sfO at.your.service say.PFV-3sfS-M ‘When the woman said to her “Seble”, she said “At your service!”. (652) tɨkʼʷɨr-ɨm nəcʼə-m kabort yə-txəttər-əma gɨred yɨ-rotʼ-əma. black-ALSO white-ALSO coatѦ REL-dress.PFV-3pfS girls 3S-run.IPFV-pfS ‘Girls wearing black and white coats are running.’

Feminine gender only concerns humans but not animals even if their natural sex is female. Thus the seemingly strange contradiction between a chicken laying eggs and masculine subject marking in (653) is due to the masculine default gender.

(653) kʷɨtara ɨnkʼʷɨra yɨ-cʼən. chicken egg 3smS-lay.IPFV ‘Chickens lay eggs.’

When a word like tɨkə ‘child’ refers to a human being without specification for gender, verbal agreement still can distinguish between feminine (654) and masculine (655). When used non-specifically as in (656), speaking about ‘the child’ irrespective whether it is a boy or a girl, the default gender masculine is chosen.

(654) a. zɨ tɨkə am-mʷət-əc. DEM child NEG-die.PFV-3sfS ‘This child () did not die.’ b. yə-z tɨkə fəzəz-ə-na-m. DAT-DEM child get.well.PFV-3smS-3sfO-M ‘This child () got well.’ (655) a. ãfʷ yɨ-kəft-e an-səna; gəna tɨkə-w. mouth 3smS-open.IPFV-PURP NEG-arrive.PFV[.3smS] stillѦ child-COP.3smS ‘He has not yet opened [his] mouth; he is still a child (m).’ b. yə-tkə-ta fɨrank yɨ-sərkʼ-xəma amʷənə-n-ɨm. DAT-child-3smPOSS money 3smS-steal.IPFV-COMP do.PFV[.3smS]-3smO-M ‘He made his child (m) steal money. (656) tɨkə-we wəxe kʼar e-trəkʼ. child-DEF good THING NEG.3smS-grow.up.IPFV ‘The child does not grow up well.’

Further, a mixed group of males and females, be it in a coordinated phrase (657) or simply juxtaposed (658), is always treated as masculine.

209 (657) gərəd-ɨm ərc-ɨm con-o-m-tanə yɨ-djakəm-o. girl-ALSO boy-ALSO sit.PFV-3pmS-CV.M-LINK 3S-bash.each.other.IPFV-pmS ‘A girl and a boy bash each other sitting.’ (658) xʷet gɨred at ərc banə-βo. (*banə-βəma) two girls one boy BE.PT-3pmS ‘There were two girls [and] one boy.’

The default gender being masculine, it goes without saying that all inanimated objects and abstract nouns are masculine as in the following examples.

(659) zɨx sənda sɨsə kʼar y-art’. DEM knife thin THING 3smS-cut.IPFV ‘This knife cuts thin things (only).’ (660) afinjə nɨkkʼar yɨ-məkkʸɨr yə-xər-e chili very 3smS-burn.IPFV REL-become.PFV-3smS.PURP a-n-nəmʷd-ɨn. NEG-1sS-love.IPFV-3smO ‘Because chili is very hot (lit. burns), I do not like it.’ (661) tɨrama zɨraβ zənəβ-ə-m. yesterday rain rain.PFV-3smS-M ‘Yesterday it rained.’ (662) amədar kʼʷəttʼər-ə-n-ɨm. cold kill.PFV-3smS-3smO-M ‘The cold (weather) killed him.’

4.2.3.1 Feminine derivational morpheme -wət The morpheme -wət occurs on some nouns referring exclusively to females, for example gəmbənwət ‘dark one ()’ or mərkamwət ‘beautiful one ()’.¹⁹⁴ With eβər- ‘so-and-so’ the feminine -wət contrasts with -yə¹⁹⁵ that stands for masculine, i.e. eβərwət ‘so-and-so ()’ vs. eβəryə ‘so-and-so (m)’. Consider its use in the following sentence:

(663) bərdəfərə cʼərəto, agaz mʷani, agaz sərəto, agaz eβəryə, abbagada sɨla, berdefere Ch. agaz M. agaz S. agaz so.and.so.m abbagada S. ʃe muxəmməd yɨ-wr-ɨyo yɨ-wr-ɨyo səβ ɨnnɨm sheikh M. 3S-say.IPFV.IPS-3pmO 3S-say.IPFV.IPS-3pmO person all bə-təra-xʷna mɨker yɨ-sətʼ-o. LOC-turnѦ-3pmPOSS monthly.gathering 3S-drink.IPFV-pmS ‘The men called berdefere Chereto, agaz Mʷani, agaz Sereto, agaz so- and-so, abbagada Sila, sheikh Muhemmed all in their turn celebrate (the monthly gathering) miker.’

¹⁹⁴Banksira (2000: 218f.) calls -wət feminine superlative, for example mərkamwət ‘the most beautiful ()’. Rose (2007: 423) states that words with -wət “may [...] have a superlative meaning”. ¹⁹⁵Hetzron (1977: 53) mentiones the form eβərwə for masculine (seeminlgy for Chaha).

210 4.3 Pronouns 4.3.1 Independent pronouns Table 83 shows the independent pronouns. Like in the verbal conjugation they distinguish between first, second and third person, singular and plural number, as well as masculine and feminine gender in the second and third persons. How- ever, there is no pronoun for the Impersonal. All second persons are based on ax- with number/gender extensions. The third persons are built on xɨ which is a demonstrative element (↗ 4.3.2) with number/gender specific suffixes. Note the palatalization as a sign of feminine gender in the singular forms axʸ and xʸɨt(a) as well as the feature [+round] in the masculine forms axu, xʷɨt(a) and xɨno. The third person singular pronouns contain the dental element t which also derives from a demonstrative element known from other Semitic languages (cf. Leslau 1979c: 369). Further, the third persons singular have two variants, one with final a and one without. The difference between them is not clear, but probably the final a expresses a certain additional demonstrative meaning comparable to the contrast xɨ and xa of the demonstratives (↗ 4.3.2). This is also supported by the fact that axʸ sometimes, though very rarely, can appear as axʸa.

  1 ɨyya yɨna (~ ɨna) 2m axə axu 2f axʸ axma 3m xʷɨt ~ xʷɨta xɨno 3f xʸɨt ~ xʸɨta xɨnəma

Table 83: Independent pronouns

Syntactically, the independent pronouns function like nouns or noun phrases. They can occur as subjects (664) or objects, usually with yə- (665) (↗ 4.7.1.2), or as heads of any other adposition (666)-(668).

(664) xʷɨt bə-mʷət-ə gɨzyə axə wə-mbər enə-βxə. 3sm LOC/TEMP-die.PFV-3smS time 2sm INF-live NEG.EX[.3smS]-MAL.2sm ‘When he is [already] dead, you should not live.’ (665) y-axə an-nəgəd-kə. DAT-2sm NEG-touch.PFV-1sS.2smO ‘I did not touch you.’ (666) yə-kʸəsəs-xʷ-ɨn bə-xʷt banə. REL-accuse.PFV-1sS-3smO LOC/INSTR-3sm BE.PT.3smS ‘I accused him for this [reason].’ (667) t-axə gi a-n-an-nə. COM-2sm time NEG-1pS-go.IPFV-1pS ‘We do (will) not go with you.’

211 (668) y-ɨyya-xəma dəm-u. DAT-1s-like blood-COP.3smS ‘According to me it is blood.’

In combination with adpositions as in the last three examples (666)-(668), the independent pronouns can of course not be left out. However, Gumer being what is often called a “PRO-drop” language, pronouns are not obligatory when they are arguments of the verb. In particular in subject or object position they are overtly stated only when they are somehow ‘emphasized’ (‘focused’ or ‘topicalized’) expressing a contrast. In (669) the pronoun ɨyya ‘I’ underlines that the speaker and not someone else will pay. In (670), there is a change of subject between the subordinate and the main clause where ɨyya ‘I’ contrasts with xʸɨta ‘she’ and as the new subject represents the new ‘topic’.

(669) ʃərət-m-axə dəmoz-m-axə ɨnnɨkʼar-axə ɨyya food-ALSO-2smPOSS salary-ALSO-2smPOSS everything-2sPOSS 1s ə-kəs-te bʷar-ə-n-ɨm. 1sS-pay.IPFV-FUT.DEF say.PFV-3smS-3smO-M ‘He said to him: “Your food and your salary, I will pay everything of you.”’ (670) təxankʼʸə lɨk xʸɨta tɨ-t-ar ɨyya yɨ-trəs-e. afterwards justѦ 3sf TEMP-3sfS-go.IPFV 1s 3smS-forget.IPFV-1sO ‘Then, just when she goes, I forget it.’

Having a demonstrative origin, the third person forms also function as definite markers suffixed to the noun they determine. For their use refer to section 4.4.2.

(671) a. ərc-xʷɨt(a) boy-DEF.sm ‘the boy’ b. dengʸa-xɨno boys-DEF.pm ‘the boys’ c. mɨʃt-xʸɨt(a) woman-DEF.sf ‘the woman’ d. ɨʃta-xɨnəma women-DEF.pf ‘the women’

4.3.2 Demonstratives Gumer possesses two demonstrative bases, z- and x-. Both occur as basic forms zɨ and xɨ (the only words in Gumer that end with the epenthetic vowel ɨ) and with a as za and xa. Further, all forms except xa can be augmented by -x, the number of possible demonstrative forms amounting to seven. As it seems, the forms with and without -x are interchangeable in most contexts.

212 etradmr opeepcuefrhrivsiain r edd h follow- a The gain needed. To are usages. investigations examples endophoric do further and ing forms picture exophoric different complete in the more often things, and equally better other occur Among clear. to less note seem to is not important here is situation It the references. oc- also text-internal that forms making all usages addition, endophoric In in world. extra-linguistic cur the in entities at pointing ing (672) entities. h aesronig).I n case, any In surroundings)’. same with the away forms far the is that something seems it that Sometimes say to addressee, sense the make to closer not but does it Since clear. entirely not drseae h ifrnebtentetodsa demonstratives distal and two speaker the where from between apart out difference set clearly somewhat The are is are. that they addresse space Usually a addressee. in and located and speaker reach demonstratives from of the both far contrast, within are vicinity In that relative sense. places the in broad located a still in but speaker reach the from away further bit a eiladtodsa eosrtvs hr sabsccnrs ewe h two the between contrast basic a with is There forms demonstratives. distal two and medial a Table 84: Table ⁹I em htmn te uaevreisdsigihls hnfu ees codn to According levels. four than less distinguish varieties Gurage other many that seems open. ¹⁹⁶It use their for reason exact the leaving narratives from taken sentences zɨ ⁹C.Amharic ¹⁹⁷Cf. nomn eot hti h egbrn Gyeto neighboring the in that reports informant hc sa niainta yt lacks Gyeto that indication an is which ( sna h speaker, the near is 2011 FAR NEAR ofr h xmlsdsrbdtedmntaie neohrcuae involv- usages exophoric in demonstratives the described examples the far, So 84 20. hyuulyhv iaydsicinproximal distinction binary a have usually they 1240f.) :      abet xa bet za(x) bet xɨ(x) bet zɨ(x) Demonstratives ugssta ue a four-way a has Gumer that suggests ɨ yɨh xrsig‘er s h w om with forms two the vs. ‘near’ expressing (673) ti’vs. ‘this’ ta os oe hr,ee ate,teohroeoe there)’ over one other the farther, even there, (over house ‘that there)’ (over addressee)’ house the ‘that to next there, (right speaker)’ house the ‘this/that to next here, (right house ‘this xɨ xa zɨ za - (676) ~ ~ ~ xɨ zɨx xɨx ya zax a ecaatrzda erteadeseo generally or addressee the near as characterized be can lutaeteocrec falfu eosrtvsin demonstratives four all of occurrence the illustrate ta’( ‘that’ xa ahrrfr oapaeta sfrhraa than away farther is that place a to refers rather elu1979c Leslau xɨ . xa 213 osnt o xml,rfrt non-visible to refer example, for not, does x 701). : ol edsrbda teohroe(in one other ‘the as described be could za ¹⁹⁶ sue uhmr fe hni Gumer, in than often more much used is ytmcnitn faproximal, a of consisting system zɨ(x) a xrsig‘far’. expressing za s distal vs. and xa za(x) ohrfrto refer both za ieie an Likewise, . and ¹⁹⁷ While Meyer xa za is . (673) ɨyya tə-z-m ema ə-wərd-e ə-ʃə banə. 1s ABL-DEM-ALSO way 1sS-go.down.IPFV-PURP 1sS-want.IPFV AUX.PT ‘I wanted to go down this (the same) way.’ (674) ɨruz tɨ-cəkkɨr-wə bə-x ɨxa. rice 3sfS-cook.IPFV-MAL.3sm INSTR-DEM water ‘She cooks rice with it, with this water.’ (675) zax aβəʃa-xʷɨta kʼɨrr bʷar-ə-n-ɨm. DEM Habasha-DEF.sm IDEO.angry say.PFV-3smS-3smO-M ‘That Habasha got angry.’ (676) xa səβ alaβa-w. DEM person A.-COP.3smS ‘That person is (an) Alaba.’

As it is the case with the demonstratives in exophoric use, there does not seem to be a considerable difference between the alternate forms with (677) and without (678) -x.

(677) ɨnnɨmkʼar arwəʃʃə-n-ɨm, zɨ aβəʃa-xʷɨta. everything teach.PFV[.3smS]-3smO-M DEM Habasha-DEF.3sm ‘He taught him everything, this Habasha.’ (678) bə-mʷət-ə gɨzyə zɨx ərc səβɨr bəkkər-ə-m. LOC/TEMP-die.PFV-3smS time DEM boy patience lack.PFV-3smS-M ‘When he died, this boy lost patience.’

Like nouns, the demonstratives can obtain nominal morphology as the illustrative choice of examples with adpositions (679) and possessives (680) show.

(679) yə-z ‘of this’ bə-x ‘with that’ tə-za ‘from that’ xɨx-e ‘to that, there’ bə-zx-e ‘in this, here’ xɨx-e-nyə ‘towards that, to there’ (680) zɨx-əta ‘this one’ xɨx-əta ‘that one’ za-ta ‘that one’ xa-ta ‘that one’ zɨx-əxʷna ‘these ones’ xa-xʷna ‘those ones’

There are two formal points to note concerning zɨ and xɨ. First, after prepositions they lose (or rather do not have to have) the final vowel ɨ (681); second, before suffixes only the forms zɨx and xɨx with the additional -x are possible (682).¹⁹⁸

¹⁹⁸Note that a form like təze derives from tə-za-y ( ~ tə-za-e).

214 (681) a. bə-z mədər (*bəzɨ) LOC-DEM place ‘at this place’ b. bə-x ɨxa (*bəxɨ) INSTR-DEM water ‘with that water’ (682) a. zɨx-əxʷna (*zəxʷna) DEM-3pmPOSS ‘these ones’ b. xɨx-əta (*xəta) DEM-3smPOSS ‘that one’

The use of demonstratives and definite articles do not exclude each other. As can be seen in the first word of (683), they even can attach directly to each other without overt head noun.

(683) tə-za-xʷɨta t-i-βr-o t-i-βr-o samt COM-DEM-DEF.sm TEMP-3S-say.IPFV-pmS TEMP-3S-say.IPFV-pmS week b-əkkəs-o gɨzyə zax xɨnd-xʷɨta y-aβəʃa kʼʷankʼʷa LOC/TEMP-wait.PFV-3pmS time DEM Indian-DEF.sm ATTR-Habasha languageѦ gəppʷa-n-ɨm. enter.PFV[.3smS]-3smO-M ‘Speaking and speaking with that one, when they waited one week, that Indian understood the language of the Habasha (i.e. Amharic).’

The same is true for possessives in (684)-(685) or (680) above.

(684) za ɨxa-ta tɨ-ʃə-n-xəma an-xar-xʷ. DEM water-3smPOSS 3sfS-want.IPFV-3smO-COMP NEG-know.PFV-1sS ‘I did not know that she wants that water.’ (685) zɨx-əta e-trəss-e. DEM-3sPOSS NEG.3smS-forget.IPFV-1sO ‘I will not forget this.’

4.3.2.1 The demonstrative element -kk- There is another demonstrative element -kk- with a general meaning ‘like (this)’. On the one hand, it appears in ɨkkɨm ‘just, simply’ (686), on the other hand there is the form ɨkka ‘like this, such, thus’ (687).

(686) səβ enə-wə, bet enə-wə, nɨβrət person NEG.EX[.3smS]-MAL.3sm house NEG.EX[.3smS]-MAL.3sm living enə-wə, ɨkkɨm meda-w. NEG.EX[.3smS]-MAL.3sm just field-COP.3smS ‘There is no person (in it), there is no house (in it), there is no life (in it), it is just land.’

215 (687) zɨx-əta yənkʼar ɨkka bar-o-n-ɨm? DEM-3smPOSS why like.this say.PFV-3pmS-1sO-M ‘Why did they say this to me like this?’

Both can occur with preceded demonstratives z- or x- as zɨkkɨm / xɨkkɨm (688) and as zɨkka / xɨkka (689)-(690). These forms can occasionally also be used as spacial adverbs (see below).

(688) t-ɨngʷəd aβəʃa gɨzyə zɨkkɨm yə-wəndɨmu-xəma COM-other Habasha time like.this DAT-W.-like ‘together with another Habasha, just like Wendimu’ (689) əgi zɨkka yə-xɨr. okay like.this 3smS-become.JUS ‘Okay, it shall be like this.’ (690) xɨkka tʼəwətʼ-əc-n-ɨm. like.that take.PFV-3smS-3smO-M ‘She held it like that.’

4.3.3 Further occurrences of the demonstrative elements 4.3.3.1 Presentatives The demonstrative elements zɨ and xɨ also occur in presentative expressions. As an independent word (or one-word predication), zɨ means ‘here (it is), take’, mostly said when handing over something to someone. In a copula sentence with either zɨ or xɨ there is an inserted geminate mm before the copula -u ‘is’. They are used usually used when pointing at something. Finally, there is zɨmʷɨnnyə which is used mostly when one finds an item after looking for it (as an answer to ɨndem-u ‘where is it?’, ↗ 4.3.5). However, as it seems these expressions interchangeable to a certain degree.

(691) zɨǃ ‘Here you are! Take!’ (mostly when handing over s.th.) zɨ-mm-uǃ ‘Here it is!’ (mostly when pointing) xɨ-mm-uǃ ‘There it is!’ (mostly when pointing) zɨmʷɨnnyə! ‘Here it is!’ (mostly when finding something)

4.3.3.2 Spatial adverbs Spatial adverbs (‘here’, ‘there’) are formed with the demonstratives and (usually) one or both of the local affixes bə- or -e. Generally speaking, the suffix -e is used for movements (692) and the prefix bə- (with or without -e) for positions (693), but other constellations are also quite frequent as in (694)-(696).

(692) zɨx-e ne-xəǃ DEM-GOAL come.IMP-2smS ‘Come hereǃ’

216 (693) yə-xno bə-z-ɨm te-tə-x-no-tanə… DAT-3pm LOC-DEM-ALSO leave-CV.T-1sS-3pmO-LINK ‘Leaving them (exactly) here…’ (694) bə-za b-attər-o gamʷə… LOC-DEM LOC/TEMP-spend.the.night.PFV-3pmS time ‘When they spent the night there…’ (695) bə-zx-e nəppər-xʷ-ɨm. LOC-DEM-GOAL live.PFV-1sS-M ‘I lived here.’ (696) xɨx-e nər-ə. DEM-GOAL EX-3smS ‘It is there.’

In addition, zɨkka ‘like this’ (↗ 4.3.2.1) can be employed to express spatial relations even without further affixes (697). Also consider the spatial expressions with zɨkka, xɨkkɨm and zɨkkɨm¹⁹⁹ plus prepositions in (698)-(699).

(697) zɨkka awra-n. like.this put.IMP[.2smS]-3smO ‘Put it here.’ (also: ‘Put it like this.’) (698) bə-zɨkka nəkəβ-xʷ-ɨn-ɨm. LOC-like.this find.PFV-1sS-3smO-M ‘I found it here.’ (699) wərəjə-ta xʷet əcʼɨr-u, tə-zɨkkɨm at-u, space.in.front.of.house-3smPOSS two fence-COP.3smS ABL-like.this one-COP.3smS tə-xɨkkɨm at-u. ABL-like.that one-COP.3smS ‘The wereje is [defined by] two fences, one from here, one from there.’

Moreover, demonstratives can combine with other morphemes like axɨr ‘about, as much as’ and gi ‘time’, sometimes with the focusing element -m. They correlate with the question words mɨraxɨr ‘how much’ and mɨrgi ‘when’ (↗ 4.3.5).

(700) z-axɨr ‘this much’ x-axɨr ‘that much’ xɨ-m-axɨr ‘that much’ xɨ-m-gi ‘just then, at once, (that time)’

4.3.3.3 Clause conjoining with xɨ and zɨ The demonstratives zɨ and especially xɨ appear abundantly in clause conjoining comparable to English ‘and then’ or, reflecting their demonstrative origin, ‘after this/that’. On the one hand, xɨ (and sometimes zɨ) are used as independent words,

¹⁹⁹I do not have any examples with xɨkka.

217 on the other hand they occur in combination with adpositions. As for the latter, the most frequent form is tə- ‘from’ + x- ‘that’ + ank’ʸə ‘after’, which can further be extended by -e ‘, to’ or -əta ‘3sm’. Further, but less frequent, the forms can also be built with the prefixes bə- ‘’ and yə- ‘; ’ and the demonstrative z- ‘that’ in any combination.

tə- x- [-e] bə- ankʼʸə [ ]²⁰⁰ yə- z- -(ə)ta

Figure 8: ‘afterwards, and then’

The following two examples (701) and (702) illustrate their clause conjoining use.

(701) yə-mɨryəm səβ nɨ-cot-nə-m nɨ-βəra-nə banə. DAT-whatever person 1pS-work.IPFV-1pS-CV.M 1pS-eat.IPFV-1pS AUX.PT təxankʼʸe ləwtʼ cənə-m. afterwards changeѦ come.PFV[.3smS]-M ‘We used to eat (i.e. live) working for whoever. Then the (political) change happened.’ (702) kʷɨtara-ta awəttʼa-xʷ-ɨn-ɨm. bəxankʼʸə ne-xʸ chicken-3smPOSS take.out.PFV-1sS-3smO-M afterwards come.IMP-2sfS kʷɨtara-xʷɨta bəssər-ə-m ba-x-na-m. chicken-DEF.sm cook.PFV-3smS-M say.PFV-1sS-3sfO-M ‘I took out the chicken. Then I said to her “Come, the chicken is cooked”.’

Note that the demonstrative element occasionally misses as in y-ankʼʸ-e (703).

(703) yankʼʸe at bet nəkəw-ni-m wər-xʷ-ɨm. afterwards one house find.PFV.IPS-BEN.1s-CV.M go.PFV-1sS-M ‘Then they found me a house [and] I went.’

Further, xɨ (and zɨ) alone also function as clause conjoining items. Often their use is combined with a short (hesitating) speech pause or lengthening of the vowel, probably comparable to English ‘so…’. Consider the following three examples (704)-(706), in particular (706) where xɨ is followed directly by another demonstrative, za, which shows that xɨ here indeed is not a demonstrative that determines a noun but serves the function of clause conjoining.

(704) bə-βəkan ɨncʼɨm-əxʷna y-ar-əβo, yə-xʷɨjɨr-əxno tʼatʼa LOC-B. naked-3pmPOSS 3S-go.IPFV-pmS DAT-clothes-3pmPOSS troubleѦ enə-no. xɨ... begi yɨ-wr-i mədər wən-nə-m. NEG.EX[.3smS]-3pmO DEM B. 3S-say.IPFV.IPS-3smO place go.PFV-1pS-M ‘In Bekan they walk around (lit. go) naked, they do not care about their clothes. Then... we went to a place called Begi.’

²⁰⁰There is an alternative form with a different final vowel -(ə)tə.

218 (705) kɨtfʷə acənəw-i-m... zɨ... dɨrə barɨya yɨ-rəβɨr kitfo bring.PFV.IPS-3smO-M DEM formerly slave 3smS-live.IPFV an-nəppər-ə? NEG-live.PFV-3smS ‘Kitfo was brought – so… weren’t there slaves in the past?’ (706) xɨ… za fərənji tə-kʼʸɨn-kʼʸɨn-əta n-an-nə, yɨna. DEM DEM foreigner ABL-buttocks-buttocks-3sPOSS 1pS-go.IPFV-1pS 1p ‘So… that foreigner, we follow him, we.’

4.3.4 Indefinite pronouns The indefinite pronouns mʷanim ‘any’, mɨryəm ‘whatever’ and attɨm ‘no; any’ are formed with the suffix -m added to mʷan ‘who’, mɨr ‘what’ and att- respectively. The former two additionally feature an inserted i or yə with unknown origin and function. The latter is related to the numeral at ‘one’ but differs from it having a geminate tt.²⁰¹ It also occurs without the nasal as attɨkʼar. While they can stand alone, often the indefinite pronouns occur in so-called series followed by one of the specifying words səβ ‘person’, kʼar ‘thing’, mədər ‘place’ or gamʷə ‘time’ (cf. Haspelmath 1997: 21).

(707) mʷanim ‘any’ mɨryəm ‘whatever’ attɨm ‘no; any’²⁰² ‘person’ mʷanim səβ mɨryəm səβ attɨm səβ ‘thing’ mʷanim kʼar mɨryəm səβ attɨnkʼar~ attɨkʼar²⁰³ ‘place’ mʷanim mədər mɨryəm mədər attɨm mədər ‘time’ mʷanim gamʷə mɨryəm gamʷə attɨm gamʷə

Typically, mʷanim and mɨryəm occur in affirmative (708)-(711) contexts, whereas attɨm combines with negated verbs (712)-(715).

(708) mʷanim gamʷə bə-gʷad-ə-n […] any time COND-be.hungry.PFV-3smS-3smO ‘Anytime when one is hungry […]’ (709) əssət-axə bə-mʷanim mədər kʼɨβr-ot tɨ-cɨl. enset-3smPOSS LOC-any place plant-INF 2smS-canѦ.IPFV ‘You can plant your əssət anywhere.’ (710) yə-mʷanim-əta ba-xʷ-ɨm xʷet bəna-xʷ-ɨm. DAT-any-3smPOSS say.PFV-1s-M two eat.PFV-1sS-M ‘In any case, I ate two.’

²⁰¹The pronounciation of the indefinite pronoun attɨm clearly contrasts with at-ɨm ‘also one, even one, and one’, i.e. at. ²⁰²With negative verb often to be translated with ‘no…’, otherwise ‘any…’. ²⁰³In accordance with Leslau (1979c: 105) written together.

219 (711) b-acənəw-i mɨryəm səβ yɨ-ttʼəkkʼəm banə. COND-bring.PFV.IPS-3smO whatever person 3smS-take.advantage.IPFV AUX.PT ‘When one brought [them], anyone took advantage. (712) zɨx-əta ɨyya bɨ-n-amʷe-n attɨm enə-wə. DEM-3sPOSS 1s TEMP-1sS-do.IPFV-3smO any NEG.EX[.3smS]-MAL.3sm ‘When I do this, it is no problem.’ (713) ɨntəganə sera enə-n, t-attɨm gənə. Endegeñ custom NEG.EX[.3smS]-3smO COM-any country ‘The Endegeñ do not have [shared] customs (traditions), with no country.’ (714) attɨm səβ t-e-aʒ-o […] any person TEMP-NEG.3S-see.IPFV-pmS ‘Before anybody sees²⁰⁴ them, […]’ (715) attɨkʼar e-xʸɨr soresa-w. anything NEG.3smS-know.IPFV hero-COP.3smS ‘He is a hero that does not know (i.e. worry about) anything.’

Finally, note that in local indefinite expressions the specifying word mədər ‘place’ often is not necessary.

(716) bə-mʷanim enə. LOC-any NEG.EX[.3smS] ‘He is nowhere.’ (717) attɨm-e wər-ot an-ʃə. any-GOAL go-INF NEG-1sS-want.IPFV ‘I do not want to go anywhere.’

4.3.5 Question words Formally, the question words of Gumer can be divided into two types: basic and composed or derived. In table 85 the latter ones are shown indented following the question word they are based on. Synchronically, mɨkʼar ‘what’ and especially məmɨr ‘how’ could possibly also be counted as basic words even though they derive from mɨr ‘what’. In all other cases the derivational morphemes are still quite transparent. In particular the prefixes yə-, bə- and tə- function basically the same way as with nominals (↗ 4.7). As for məcə and məcra ‘when’, the latter is represented as the derived form due to its additional ending -ra.²⁰⁵ The question words meaning ‘where’ are based on e, but (probably) more frequently one finds ete.²⁰⁶ It is not entirely clear whether its final -e is connected to -e ‘; ’ (↗ 4.7.2) or not. As for et ‘which’, note that it should be kept apart from ete.

²⁰⁴Usually the Imperfective verb base in 2/3p is t-e-az-o with voiced z rather than ʒ, cf. 3.16. ²⁰⁵In other Semitic languages the basic word for ‘when’ does not contain -ra or similar, for example Arabic matā or Amharic məce (Leslau 1979c: 387). ²⁰⁶According to Leslau (1979c: 1), e ‘where’ exists only in Chaha (and by extension in Gumer), the other Gurage varieties (but also including Chaha) featuring ete or a corresponding form.

220 mʷan ‘who’ yəmʷan ‘whose’; ‘for who’ < yə- ‘’; ‘’ + mʷan ‘who’ mɨr ‘what’ mɨkʼar ‘what’ < mɨr ‘what’ + kʼar ‘’ yənkʼar ‘why, what for’ < yə- ‘’ + mɨkʼar ‘what’ məmɨr ‘how’ < mɨr-mɨr ‘what-what’²⁰⁷ yəmɨr ‘why’ < yə- ‘’ + mɨr ‘what’ mɨrgi ‘when (what recent time)’ < mɨr ‘what’ + gi ‘time’ mɨraxɨr ‘how much, how many’ < mɨr ‘what’ + axɨr ‘equal’ mɨraxɨrgi ‘how many times’ < mɨraxɨr ‘how many’ + gi ‘time’ mɨr enət ‘what kind’ < mɨr ‘what’ + enət ‘kind’ mɨrina ‘what language’ < mɨr ‘what’ + -ina²⁰⁸ məcə ‘when’ asking about the future, with  məcra ‘when’ asking about the past, with  e~ ete ‘where (to)’ be~ bete ‘where’ < bə- ‘’ + e(te) ‘where’ tete ‘where from’ < tə- ‘’ + ete ‘where’ etenyə ‘in which direction’ < ete ‘where’ + -nyə ‘towards’ et(-) ‘which’ ɨndem- ‘where (is)’ said when s.th. should be around

Table 85: Question words

The following two examples show typical short questions consisting of a question word followed by the verb.

(718) mʷan yə-mbər? who 3smS-live.JUS ‘Who should live?’ (719) ete-n-xʸ? where-COP-2sfS ‘Where are you (s)?’

The most common position of the question words is directly before the verb, no matter if it is the subject (720) or object (721).

(720) zɨ wɨssa mʷan aβəssər-ə-m? DEM wussa who cook.PFV-3smS-M ‘Who baked this wussa-bread?

²⁰⁷According to Leslau (1979c: 406) məmɨr derives from mɨr-mɨr. ²⁰⁸See example (605).

221 (721) luβaβat mɨr aβəssər-əc-ɨm? L. what cook.PFV-3sfS-M ‘What did Lubabat bake?’

Also in more complex question with subordinate clauses the question words normally occur before the verb of the clause it belongs to rather than at the beginning of the whole sentence.

(722) zɨxe-nyə yɨ-rəx-e mɨr yɨ-xʷəra-n? here-DIR 3smS-send.IPFV-PURP what 3smS-prevent.IPFV-3smO ‘What does prevent him to send [someone] here?’ (723) bə-fətəna bə-wətʼəkʼ-xə mɨr t-ame-te? LOC-exam COND-fall.PFV-2smS what 2smS-do.IPFV-FUT.DEF ‘What will you do if you fail in the exam?’

Further, consider also the question word in (724) appearing in a dependent clause (quoted speech) and preceding the verb denoting the main content, regardless of whether this is an infinitive or whether there are more verbal elements following or not.

(724) əxʷa ɨyya mɨr wə-tot nər-ə-βi ba-xʷ-ɨm assəβ-xʷ-ɨm. now 1s what INF-make EX-3smS-MAL.1s say.PFV-1sS-M think.PFV-1sS-M ‘I thought: “Now, what do I have to do?”’

Nevertheless, it is also quite possible for a question word to appear in a different position like, for example, at the beginning of a sentence before the object (725).

(725) yəmɨr ɨxa-xʷɨta xʷə-xʸ-ɨm bar-əc-ɨm. why water-DEF.sm spill.PFV-2sfS-M say.PFV-3sfS-M ‘She said: “Why did you spill the water?”’

One possible reason for this must be information structure, but also the type of question word might play a role. Other than mʷan ‘who’ and mɨr ‘what’ in most examples above, yəmɨr ‘why’ does not function as a core argument of the verb. In particular mɨr and mɨkʼar ‘what’ often ask for the object which generally tends to occurs right before the verb anyway (726).

(726) ogʸət, mɨkʼar t-aʒ? O. what 2smS-see.IPFV ‘[People o] Ogyet, what do you think (lit. see)?’

Since the thing or person one asks about constitutes the focus of a sentence, question words appear quite often in cleft sentences. In this case their position can change to the front or the end of a sentence. However, this is only apparent when there are enough other words intervening between question word and verb as in examples (727)-(734).

222 (727) mʷan-u yɨ-dəwwɨl? who-COP.3smS 3smS-phoneѦ.IPFV ‘Who is calling?’ (728) təxankʼʸə bə-jəppər-xu ankʼʸə mɨkʼar-u afterwards LOC/TEMP-finish.PFV-2pmS after what-COP.3smS yə-tənəf-ə wə-βər-u. REL-remain.PFV-3smS INF-say-COP.3smS ‘And then, after you have finished, what’s then (left to do), actually?’ (729) məmɨr-u y-amʷər-i? how-COP.3smS 3smS-do.IPFV.IPS-3smO ‘How is it done / how do they do it?’ (730) t-ete-w yə-səna-xə? ABL-where-COP.3smS REL-arrive.PFV-2smS ‘Where did you arrive from?’ (731) bə-nkʼar-u əxʷa təwag-ot-əta? LOC-what-COP.3smS now fight-INF-3smPOSS ‘About what is the fighting now?’ (732) yə-nkʼar-u yɨ-dəkʼ? DAT-what-COP.3smS 3smS-laugh.IPFV ‘Why is he laughing?’ (733) əxʷa bə-mɨr-u səβatənə fokʼ ə-wətʼa? now INSTR-what-COP.3smS seventh floorѦ 1sS-go.up.IPFV ‘Now, with what do I go up to the seventh floor?’ (734) mʷan-u za-ta nəgə cənə-m yɨ-cot-ɨn? who-COP.3smS DEM-3smPOSS tomorrow come.PFV[.3smS-M] 3smS-make.IPFV-3smO ‘Who can come to/and do this tomorrow?

Question words can appear doubled when they are meant to refer to more than one entity. In example (735), the speaker uses the reduplicated mʷan-mʷan ‘who (pl)’ to enquire about more than one person, presumably intending to hear all individual names. A slightly different case is the doubled question word mɨraxɨr ‘how much’ in example (736). Here it is used in a distributive meaning expressing the notion of ‘each’.

(735) mʷan-mʷan yɨ-cən-te bə-gən? who-who 3smS-come.IPFV-FUT.DEF LOC-country ‘Who (pl) will come from the countryside?’ (736) mɨraxɨr mɨraxɨr aβ-o-ku-m? how.much how.much give.PFV-3pmS-2pmO-M ‘How much did they give you each?’

The usage or meaning of some of the question words does not always correspond one-to-one to the English translation. For example, when asking for someone’s ʃɨm ‘name’, mʷan ‘who’ is used rather than məmɨr ‘how’ or mɨr ‘what’ (737)-(738).

223 Further, to ask for the time mɨraxɨr ‘how much’ is used (739).

(737) y-axə ʃɨm mʷan nɨ-βər? ATTR-2sm name who 1sS-say.JUS ‘How (lit. who) should I call you?’ (738) ʃɨm-əta mʷan yɨ-wr-i? name-3smPOSS who 3smS-say.IPFV.IPS-3smO ‘How (lit. who) is he called? What’s his name?’ (739) sat mɨraxɨr-u? hour how.much-COP.3smS ‘What time is it?’

The difference between the two question words meaning ‘what’, mɨr and mɨkʼar, is not entirely clear. Usually they seem to be interchangeable as it is the case in (740) where the speaker used both words in what appears to be the exactly same context.

(740) əxʷa za-ta mɨr yɨ-cot, zɨx-əta mɨkʼar yɨ-cot, now that-3smPOSS what 3smS-make.IPFV this-3smPOSS what 3smS-make.IPFV mɨr yɨ-cot-o yɨ-βɨn-no. what 3S-make.IPFV-pmS 3smS-say.IPFV-3pmO ‘He says to them: “Now, what does that one do, what does this one do, what do they do?”’

The conceivable assumption that mɨkʼar, which contains kʼar ‘thing’, could refer to objects and mɨr to abstract concepts does not prove correct. Compare the following two examples which both ask for concrete objects, once with mɨr (741) and once with mɨkʼar (742).

(741) mɨr sɨyə-xə-wə-m? what buy.PFV-2smS-MAL.3sm-M ‘What did you buy with it?’ (742) mɨkʼar aβ-o-ku-m? what give.PFV-3pmS-2pmO-M ‘What did they give you?’

The question word mɨraxɨr ‘how much, how many’ is normally followed by a noun it determines as in (743), but it can also be used elliptically as in (744).

(743) mɨraxɨr kərə əkkəs-xu-m? how.many day wait.PFV-2pmS-M ‘How many days did you (pm) stay?’ (744) səβat bet gʷɨrage t-i-wr-i mɨraxɨl-l-o? seven house Gurage TEMP-3S-say.IPFV.IPS-3smO how.many-COP-3pmS ‘Speaking about Sebat Bet Gurage, how many are they?’

224 Also mɨr enət ‘what kind’ normally is followed by a noun as in (745).

(745) mɨr enət ʃərət tɨ-rəmd? what kind food 2smS-love.IPFV ‘What kind of food to you like?’

To ask ‘where’, there are two pairs of forms: e ~ ete on the one hand and be ~ bete with the locative preposition b(ə)- on the other hand. Usually the forms without b(ə)- occur with verbs of motion indicating a direction, i.e. ‘where to’ (746)-(747), whereas with the preposition they refer to locations ‘where’ (748)-(749).

(746) gaʃʃe-na e wər-ə-m? father-1sPOSS where go.PFV-3smS-M ‘Where did my father go?’ (747) ete nɨ-tʼa? where 1sS-go.up.JUS ‘Where should I go up?’ (748) ʃɨm-ʃɨm tʼənə-m yɨ-rəʃe kʼar be nər-ə? name-name call.PFV.IPS-CV.M 3S-lift.IPFV.IPS.3smO THING where EX-3smS ‘Where does such a thing exist that one distributes [food] by calling name by name?’ (749) bete nəkəβ-xʷə-n-ɨm? where find.PFV-2smS-3smO-M ‘Where did you find it?’

Nevertheless, in some cases one can also use ete instead of bete. It seems that this is only possible when a certain “movement” is involved in the sense that an entity is not anymore at a specific location but “has moved away” from it.

(750) əxʷa ete y-anə-xəma a-n-xʸɨr. now where REL-EX-COMP NEG-1sS-know.IPFV ‘I don’t know where it is now.’

Finally, e(te) can also be followed by postpositions like dar ‘until’ in (751) or, to explicitly express ‘in direction o’, by the affix -nyə ‘towards’ (752).

(751) gɨnɨɲɲunnət-ənda e dar-u? meetingѦ-1pPOSS where until-COP.3smS ‘Until where is our meeting [point]?’ (752) tɨrama ete-nyə t-ar banə? yesterday where-DIR 2smS-go.IPFV AUX.PT ‘Where were you going yesterday?’

When searching for something in the immediate context or environment but, contrary to expectations, is not findable at the moment, one uses ɨndem (plus copula) rather than bete.

225 (753) ʒɨβangʸɨβa ɨndem-u? zhibangyiba where-COP.3smS ‘Where is the zhibangyiba?’²⁰⁹

Similarly, ɨndem is also used when one is in the vicinity of several possible referents but does not know which one(s) of them are meant. Thus, for example, standing in a village with all houses within sight, one could utter the question in (754), or seeing a big group of girls one might ask the question in (755). In contrast, be(te) ‘where’ is used when generally inquiring the location of an entity “in the world”. Note that in these cases the English translation is rather ‘which’ than ‘where’.

(754) bet-axə ɨndem-u? house-2smPOSS where-COP.3smS ‘Which/where is your house?’ (755) gɨred-axə ɨndem-l-əma? girls-2smPOSS where-COP-3pfS ‘Which/where are your daughters?’

Gumer distinguishes between two question words for ‘when’, məcə and məcra. The former asks for a time point in the future (or present) and combines with the Imperfective (756)-(757). The latter is used for the past and occurs with the Perfective (758).

(756) ɨyya-m axə-m gʷəmarə məcə-w nɨ-fəka-nə? 1s-ALSO 2sm-ALSO Gumer when-COP.3smS 1pS-leave.IPFV-1pS ‘Me and you, when do we leave for Gumer?’ (757) məcə-w tɨmɨrt tɨ-kʼərs? when-COP.3smS studying 3sfS-begin.IPFV ‘When does she begin to study?’ (758) məcra-w tɨmɨrt yə-k’ənəs-xə? when-COP.3smS studying REL-begin.PFV-2smS ‘When did you begin to study?’

Note that the morpheme -ra also occurs with səstə, nəβatə and samtə to refer to days in the past rather than in the future.²¹⁰

²⁰⁹ʒɨβangʸɨβa is a tool used to pound the root of əssət. According to Leslau (1979c: 720) it is a loan word from Kʼabeena. ²¹⁰According to Hetzron (1977: 112) these terms are built on an old ordinal pattern CaCɨC, i.e. sost ‘three’ → səst- ‘third day (counted from today)’, arβət ‘four’ → nəβat- ‘fourth day (counted from today)’ (without initial vowel a the r becomes n) and sɨmʷɨt ‘eight’ → samt- ‘eighth day (counted from today)’, i.e. a week (samt also being the word for ‘week’). There seem to be further terms for fifth, sixth and seventh day.

226 (759) səstə ‘the day after tomorrow, two days from now’ səstɨra ‘the day before yesterday, two days ago’ nəβatə ‘three days from now’ nəβətra ‘three days ago’ samtə ‘a week from now’ samtɨra ‘a week ago’

The question word et ‘which’, not to be confused with e(te) ‘where’, most commonly appears together with the 3sm possessive suffix -əta and without head noun as in (760) corresponding to English ‘which one’.

(760) et-əta tɨ-rəmd? which-3smPOSS 2smS-love.IPFV ‘Which one do you like?’

If there is an overt head noun, the question word et precedes it and occurs without -əta. Note that there is the copula -u attached to the head noun forming (what appears to be) a cleft sentence. However, while all other question words also occur in normal non-clefted sentences, it seems that this is impossible with et.

(761) et ʃərət-u tɨ-rəmd? which food-COP.3smS 2smS-love.IPFV ‘Which food do you like?’

Further, the copula also occurs (somewhat unexpectedly) on the head noun in sentences with predicative adjectives (762)-(763) rather than at the end of the sentence after the adjective. It is not clear what conditions this unusual word order.

(762) et bet-u tʼɨrə? which house-COP.3smS expensive ‘Which house is (more) expensive?’ (763) et gɨred-l-əma mərkama? which girls-COP-3spfS beautiful ‘Which girls are beautiful?’

4.3.6 Possessives In Gumer there are two possibilities to refer to the possessor pronominally. First, the attributivizer yə- is prefixed to the independent pronouns preceding the pos- sessum, very much like any noun that is attributed to another noun (↗ 4.7.1.1).

(764) y-ɨyya angət ATTR-1s neck ‘my neck’

227 (765) y-axə ʃɨm ATTR-2sm name ‘your (sm) name’ (766) yə-xʷta ʃərət ATTR-3sm food ‘his food’

The second and more common means are the possessive suffixes attached to the possessed noun. As table 86 shows, they distinguish person, number and gender like the independent pronouns with which they also share some formal similari- ties. The second persons, all of them beginning with a, look exactly the same as their independent counterparts. The third persons except 3sm -əta consist of ə followed by the pronoun. However, there are two freely interchangeable forms for 3pm. In the more common variant -əxʷna the feature [round] detached from the final vowel of -əxno and docked on the velar x resulting in -əxʷna. The first persons also have an initial ə. While the pronominal element of 1p corresponds to the object suffix, 1s features a completely distinct form.

  1 -əna -ənda 2m -axə -axu 2f -axʸ -axma 3m -əta -əxʷna~ -əxno 3f -əxʸta -əxnəma

Table 86: Possessive suffixes

The initial vowels of the possessive suffixes interact with preceding vowels in different ways according to the rules discussed in section 2.3.3. Various examples illustrating this are given in tables 87 and 88, in addition to their (unchanged) forms after consonants like the t of bet ‘house’. The initial a of the second persons delete a preceding ə as with tɨkə ‘child’. Both initial a and ə form hiatus with preceding o as with asso ‘salt’, and there is a glide y after i as in təmari ‘student’. In the case of tʼu ‘breast’ the vowel changes to the glide w (another possibility with u being insertion of a glide w as for example in cʼucʼuwaxə ‘your chicken’). When the same two vowels meet (i.e. ə+ə and a+a) one of them is deleted. In segmented and glossed examples the (arbitrary) convention followed here always leaves out the vowel of the suffix.

228 bet ‘house’ tɨkə ‘child’ angacca ‘cat’ 1s bet-əna tɨkə-na angacca-na 2sm bet-axə tɨk-axə angacca-xə 2sf bet-axʸ tɨk-axʸ angacca-xʸ 3sm bet-əta tɨkə-ta angacca-ta 3sf bet-əxʸta tɨkə-xʸta angacca-xʸta 1p bet-ənda tɨkə-nda angacca-nda 2pm bet-axu tɨk-axu angacca-xu 2pf bet-axma tɨk-axma angacca-xma 3pm bet-əxʷna~ -əxno tɨkə-xʷna~ -xno angacca-xʷna ~ -xno 3pf bet-əxnəma tɨkə-xnəma angacca-xnəma

Table 87: Possessive suffixes in different environments (Part I)

asso ‘salt’ tʼu ‘breast’ təmari ‘student’ 1s asso-əna tʼɨw-əna təmari-yəna 2sm asso-axə t’ɨw-axə təmari-yaxə 2sf asso-axʸ tʼɨw-axʸ təmari-yaxʸ 3sm asso-əta tʼɨw-əta təmari-yəta 3sf asso-əxʸta tʼɨw-əxʸta təmari-yəxʸta 1p asso-ənda tʼɨw-ənda təmari-yənda 2pm asso-axu tʼɨw-axu təmari-yaxu 2pf asso-axma tʼɨw-axma angacca-xma 3pm asso-əxʷna~ -əxno tʼɨw-əxʷna~ -əxno təmari-yəxʷna ~ -yəxno 3pf asso-əxnəma tʼɨw-əxnəma təmari-yəxnəma

Table 88: Possessive suffixes in different environments (Part II)

The difference between suffixed and attributed possessive pronoun is not (necessarily) neutral vs. emphasized because both can occur at the same time (767)-(768). Nevertheless, their co-occurrence might express some additional em- phasis compared to the single use of only one or the other option.

(767) y-ɨyya simkard-əna at xʷet sost arβət-u, pinkod. ATTR-1s SIM.card-1sPOSS one two three four-COP.3smS PIN.code ‘My SIM card is one-two-three-four, the PIN code.’ (768) y-axə tʼəlat-axə-w. ATTR-2sm enemyѦ-2smPOSS-COP.3smS ‘He is your enemy.’

Along these lines, an attributive phrase with two nouns (↗ 4.7.1.1) can sometimes be furnished with an additional possessive suffix on the second noun resuming the preceding attributed noun (769)-(771). However, since normally this is un-

229 grammatical as shown in (772), it remains to explore in what circumstances it is still possible.

(769) yə-səβat bet gʷɨragʷe ammətʼatʼ-əta n-od-xə. ATTR-seven house Gurage originѦ-3smPOSS 1sS-tell.JUS-2smO ‘Let me tell you about the origin of the Sebat Bet Gurage.’ (770) gʷəmarə y-atʼəfʷa-n-e b-i-cən yə-gʷəmarə Gumer 3smS-ambush.IPFV-3smO-PURP TEMP-3smS-come.IPFV ATTR-Gumer agaz-əta abbagada-ta arβa-w. agaz-3smPOSS abbagada-3smPOSS fourty-COP.3smS ‘When the Gumer come to ambush him, the agaz [and] abbagada of the Gurage are forty.’ (771) yə-gʷəmarə tə-gʸəta dar-əxʷna xʷɨt bə-βar-ə kʼar ATTR-Gumer COM-Gyeto boundary-3pmPOSS 3sm INSTR-say.PFV-3smS THING yə-xɨr. 3smS-become.JUS ‘The boundaries of Gumer and Gyeta shall be according to what he said.’ (772) *yə-fərəz wərcə-ta ATTR-horse front.leg-3smPOSS (intended: ‘the horse’s front leg)’

If required by information structure, a noun phrase referring to a possessor – presumably in particular proper names as in (773)-(774) – can be left-located into a topical position and then resumed by the possessive suffix.

(773) umər farda mʷena-ta zɨxe gʷəmarə-w. U. F. uncle.maternal.side-3smPOSS here Gumer-COP.3smS ‘Umer Farda’s uncles [and family] are here Gumer people.’ (774) ɨmɨryə tə-gəβrə_xanna adot-əxʷna ɨnjera_adot-əxʷna banə-c. I. COM-G._H. mother-3pmPOSS stepmother-3pmPOSS BE.PT-3sfS ‘Imiryeʼs and Gebre Hanna’s mother was their stepmother.’ (‘As for Imirye and Gebre Hanna, their mother was their stepmother.’)

The ‘normal’ neutral attributive construction (↗ 4.7.1.1) does not feature the possessive, thus the above sentence would be as in (775).

(775) y-ɨmɨryə tə-gəβrə_xanna adot ɨnjera_adot-əxʷna banə-c. ATTR-I. COM-G._H. mother-3pmPOSS stepmother-3pmPOSS BE.PT-3sfS ‘Imirye and Gebre Hanna’s mother was their stepmother.’

What follows are a few more illustrative example sentences showing possessive markers suffixed to nouns in various syntactic positions as object (776), local expression without adposition (777) or in a prepositional phrase (778).

(776) əgr-əxʷna yɨ-β-o-ndə. foot-3pmPOSS 3S-give.IPFV-pmS-1pO ‘They give us their feet.’

230 (777) bet-axu wər-oǃ house-2pmPOSS go.IMP-2pmS ‘Go (to your) homeǃ’ (778) y-okkʷa-c-ɨn afinjə ɨkka b-əgr-əxʸta ATTR-pound.PFV-3sfS-3smO chili like.this INSTR/LOC-foot-3sfPOSS tʼəwətʼ-əc-n-ɨm. take.PFV-3sfS-3smO-M ‘She held the afinje she had pounded with her feet (or: kept at her feet) like this.’

In particular note that the suffix attaches directly to the possessed noun and not, for example, to the adjective (779) or to the end of the whole noun phrase (780).

(779) ɨrs gʷəppay-əna fərəz yɨ-kʼyə banə. little brother-1sPOSS horse 3smS-look.after.IPFV AUX.PT ‘My little brother was looking after the horses.’ (780) yə-fərəz-əta wərcə aʃʃə-xʷ-ɨm. ATTR-horse-3smPOSS front.leg see.PFV-1sS-M ‘I saw the front leg of his horse.’

Discourse particles like -ʃ ‘you know’ follow the possessive suffix (781). In contrast, the coordinating -m ‘also, and’ exceptionally stands between noun and possessive (782). This fact suggests that the possessive suffixes are clitical rather than suffixal.

(781) dəwo-axə-ʃ an-xər-ə. relative-2smPOSS-PRAG NEG-become.PFV-3smS ‘He is not your relative (you know).’ (782) ʃərət-m-axə dəmoz-m-axə ɨnnɨkʼar-axə ɨyya food-ALSO-2smPOSS salary-ALSO-2smPOSS everything-2sPOSS 1s ə-kəs-te bʷar-ə-n-ɨm. 1sS-pay.IPFV-FUT.DEF say.PFV-3smS-3smO-M ‘He said to him: “Your food and your salary, I will pay everything of you.”’

The possessive suffixes occur in functions that usually are not viewed as possession in the traditional sense. The use of 3sm -əta as marker of (associative) definiteness is discussed briefly in section 4.4.2.1. Further, possessive suffixes of all persons occur with numerals, quantifiers and nominals like gəg ‘body’ or xʸɨn ‘heart’. In combination with numerals and quantifiers (783)-(786), the possessives express the meaning ‘ of ’.

(783) arβət-ənda four-1pPOSS ‘the four of us’

231 (784) yə-xʷecɨm-ənda awənə-ndə-m. DAT-both-1pPOSS feed.PFV.IPS-1pO-M ‘They fed both of us.’ (785) atat-ənda atat gojjo tʼəβətʼ-nə-m. some-1pPOSS some hutѦ take.PFV-1pS-M ‘Some of us took some huts.’ (786) zax səβ ɨnnɨm-əxʷna at-at-at-at-at yə-ʒɨr danə DEM person all-3pmPOSS one-one-one-one-one ATTR-zhir judge awəttʼ-o-m. take.out.PFV-3pmS-M ‘These persons, all of them […] named one zhir-judge each.’

To express reflexives (‘sel’, ‘own’), the corresponding possessives are suffixed to gəg ‘body’ (787)-(789).

(787) gʷeta gəg-əta e-trəməd, b-oxe zənga God self-3smPOSS NEG.3smS-be.expensive.IPFV INSTR-good thing yə-kʼəyə-ndə. 3smS-protect.JUS-1pO ‘God himself is not expensive, may he protect us with good things.’ (788) gəg-əta baləge yə-xər-ə-xəma yɨ-xʸɨr. self-3smPOSS stupidѦ REL-become.PFV-3smS-COMP 3smS-know.IPFV ‘He himself knows that he is stupid.’ (789) yə-gəg-əna fərəz ATTR-body-1sPOSS horse ‘my own horse’

Further, there are a few adverbial expressions formed with possessive suffixes. For example, ‘easily, jokingly’ can be expressed by yə-sya-m- on the basis of sɨya ‘joke’ (790)-(791). The construction tə-xʸn-, literally ‘from/with one’s heart’, translates as ‘slowly, carefully’ (792). Finally, also ‘naked’ always occurs with possessives as ɨncʼɨm- (793). According to Leslau (1979c: 56f.) it also means ‘only’, but it seemingly is based on the substantive or adjective ɨncʼ ‘nakedness, naked’.

(790) bɨ-n-wəgʷa-n yə-sya-m-əta xɨkka TEMP-1sS-stab.IPFV-3smO DAT-joke-ALSO-1pPOSS like.that amənə-βi-m-tanə fərəz fʷɨnkʸɨnn amənə-m. do.PFV[.3smS]-MAL.1s-CV.M-LINK horse step.aside do.PFV[.3smS]-M ‘When I stabbed him, he made easily like that (on me) and made the horse step aside a bit.’ (791) yə-βora bəsər gɨβtɨ-gɨβt yə-sya-m-ənda afettən-ne-m. ATTR-ox meat half-half DAT-joke-ALSO-1pPOSS finish.up.PFV-1pS.3smO-M ‘We easily finished up the ox meat each a half.’

232 (792) ɨnnɨmgi tə-xʸn-əta y-ar. always COM-heart-3smPOSS 3smS-go.IPFV ‘He always goes slowly/carefully.’ (793) bə-βəkan ɨncʼɨm-əxʷna y-ar-əβo. LOC-B. naked-3pmPOSS 3S-go.IPFV-pmS ‘In Bekan they walk around naked.’

4.4 Definiteness Overt marking of definiteness in Gumer is less prominent than, for example, in Romance or Germanic languages. Rose (2007: 421) even states that “[g]enerally, there is no expression of definiteness on the noun”. This does not mean that there are no morphological means to mark definiteness in Gumer. Rather it has to be understood in such a way that definite articles are required in noticeably less contexts and that they do not simply appear to signal that a noun (or noun phrase) is identifiable for the hearer (identifiability and familiarity being the main conditions triggering the use of definite articles in general, cf. Lyons 1999: 3) but that some other factors play a role as well. In addition to said definite articles, Gumer features possessive suffixes (↗ 4.3.6) and demonstratives (↗ 4.3.2), the other two commonly known categories that render nouns definite. A crucial point is the fact that the possessives not only express possession but also definiteness in contexts where other languages only use definite articles, that is Gumer distinguishes rather clearly between associative uses marked by the possessives and anaphoric uses marked by the definite articles (↗ 4.4.2.1). A further noteworthy point is that articles and possessives can combine with demonstratives (↗ 4.3.2). Finally, it is understood that the (independent) pronouns (↗ 4.3.1) and proper names are inherently definite, also without explicit additional marking for definiteness.

4.4.1 Definite articles Gumer has two types of definite articles. On the one hand, there is the invariable suffix -we which occurs rather marginally. Since it is presumably an influence from neighboring Gurage varieties like Ezha, it is addressed only briefly in 4.4.1.1 below. On the other hand there is the ‘proper’ definite article in Gumer which corresponds formally to the third person independent pronouns (↗ 4.3.1) and thus appears in four forms distinguishing number and gender as illustrated in table 89. Note that the singular articles have two forms each as do the independent pronouns. The two variants with and without final a seem to be freely interchangeable in most if not all contexts.

233   m -xʷɨt ~-xʷɨta -xɨno f -xʸɨt ~ -xʸɨta -xɨnəma

Table 89: Definite article

It is not entirely clear whether the definite article is an independent word, a clitic or an affix. Rose (2007: 422) states that it follows the noun and writes it separated, for example mɨs xʷɨta ‘the man’ or mɨʃt xʸɨta ‘the woman’. Hetzron (1977: 56), on the contrary, claims that “the appropriate third person independent pronouns are suffixed”, but in his transcribed texts it is also written independently, for example kʼərcʼat huta ‘the basket’. It seems to me that the noun and the following article form a single unit with only one intonational peak. Therefore I chose the convention to write them together as for example kʷɨtaraxʷɨta ‘the chicken’ or səβxɨno ‘the persons’ rather than kʷɨtara xʷɨta or səβ xɨno. Nevertheless, it is important to note that the definite article apparently never drops (or at least usually does not drop) the epenthetic vowel ɨ even if it were possible (or required) according to the syllable structure (↗ 2.3.2) as shown in (794)-(797). For example, in a word composed of kʷɨtara + xʷɨta the syllable boundaries are expected to be rearranged from kʷɨ.ta.ra.xʷɨ.ta to kʷɨ.ta.raxʷ.ta, but normally the output remains kʷɨtaraxʷɨta. This fact is an indication that the definite article has kept its independent status to some extent.

(794) kʷɨ.ta.ra + xʷɨ.ta → kʷɨ.ta.ra-xʷɨ.ta (*kʷɨ.ta.raxʷ.ta) chicken DEF.sm ‘the chicken’ (795) mɨʃt + xʸɨ.ta → mɨʃt-xʸɨ.ta (*mɨʃ.tɨxʸ.ta) woman DEF.sf ‘the woman’ (796) den.gʸa + xɨ.no → den.gʸa-xɨ.no (*den.gʸax.no) boys DEF.pm ‘the boys’ (797) ɨʃ.ta + xɨ.nə.ma → ɨʃ.ta-xɨ.nə.ma (*ɨʃ.tax.nə.ma) boys DEF.pf ‘the women’

The choice of the article is determined by the gender and number of the noun they are attached to. As discussed in section 4.2.3, in Gumer the default gender is masculine, a fact which makes -xʷɨt(a) the most frequent article of the four. It thus occurs with male human beings (798), animals (799), inanimate objects (800), and abstract nouns (801).

234 (798) mɨs-xʷɨta jeneral banə. man-DEF.sm general BE.PT ‘The man was a general.’ (799) kʷɨtara-xʷɨta bəssər-ə-m. chicken-DEF.sm cook.PFV-3smS-M ‘The chicken is cooked.’ (800) kʼawa-xʷɨta sɨyə-xʷ-ɨn-ɨm. coffee-DEF.sm cook.PFV-3smS-3smO-M ‘I bought the coffee.’ (801) amədar-xʷɨt baʃə-xʸta aβas-ə-βa-m. cold-DEF.sm sickness-3sfPOSS make.worse.PFV-3smS-MAL.3sf-M ‘The cold [weather] worsened her sickness.’

The only nouns that are treated as feminine are female human beings.

(802) mɨʃt-xʸɨta dakʼ-əc-ɨm. woman-DEF.sf laugh.PFV-3sfS-M ‘The woman laughed.’ (803) gərəd-xʸɨta cənə-c-ɨm. girl-DEF.sf come.PFV-3sfS-M ‘The girl came.’

The gender of the plurals remains the same as in the singular.

(804) dengʸa-xɨno əray yɨ-wər-o. boys-DEF.pm cows 3S-tend.IPFV-3pmS ‘The boys tend cattle.’ (805) ɨʃta-xɨnəma əssət yɨ-fəkʼ-əma. women-DEF.pf enset 3S-scrape.IPFV-3pfS ‘The women scrape əssət.’

The definite articles not only attach to nouns but also to other ‘nominal’ parts of speech like ʽrelative verbsʼ (↗ 4.7.3.1) as in (806)-(807), demonstratives (808), question words (809) and adjectives (809).

(806) ʃəwa yɨ-dəwwɨl-xʷɨta Addis.Ababa 3smS-phoneѦ.IPFV-DEF.sm ‘the one who calls to Addis Ababa’ (807) y-axə yɨ-kəs-xə-xʷɨt DAT-2sm 3smS-pay.IPFV-2smO-DEF.sm ‘the one who pays you’ (808) za-xɨno cənə-βo-m. DEM-DEF.sm come.PFV-3pmS-M ‘Those ones came.’

235 (809) wəmbər naməǃ – et-xʷɨt?– accʼɨr-xʷɨt. chairѦ bring.INF.2smS which-DEF.sm short-DEF.sm ‘Bring a chairǃ – Which one? – The short oneǃ’

4.4.1.1 The definite article -we As mentioned above there is a definite article -we²¹¹ that appears occasionally in Gumer. In Ezha and Muher -we is the regular definite article while in other Western Gurage languages it is not attested (Meyer 2011: 1243). Due to the fact that in Gumer it is used more often by speakers in areas closer to Ezha and Muher than to Chaha, it is quite probable that -we is an influence from Ezha or Muher. The article -we is invariable and combines with any gender and number, for example ərcwe ‘the boy’, dengʸawe ‘the boys’, gərədwe ‘the girl’ and gɨredwe ‘the girls’. In contrast to the definite article -xʷɨt, which always occurs at the end of the whole noun phrase (810), -we attaches to preceding adjectives or relative clauses that modify the noun (811).

(810) yə-rədəd-ə bet-xʷɨt ʃərətβet-u. (*yərədədəxʷɨt bet) REL-burn.PFV-3smS house-DEF.sm restaurant-COP.3smS ‘The house that burnt down is a restaurant.’ (811) yə-rədəd-ə-we bet ʃərətβet-u. (*yərədədə betwe) REL-burn.PFV-3smS-DEF house restaurant-COP.3smS ‘The house that burnt down is a restaurant.’

In recorded stories and conversations -we occurs very rarely. By contrast its frequency is slightly higher in isolated elicited sentences as in (812) and the following examples (813)-(814).

(812) mɨs-we gʷəβəz yə-xər-e gʷəncə tʼəβətʼ-ə-m. man-DEF braveѦ REL-become.PFV-3smS.PURP hyena take.PFV-3smS-M ‘Because the man is brave, he caught a hyena.’ (813) bɨkʼʷrə-we yə-tkə-xʷɨt nəkʼʷətʼ-ə-n-ɨm. mule-DEF DAT-child-DEF.sm kick.PFV-3smS-3smO-M ‘The mule kicked the child.’ (814) gʷad-ə-n-ɨm-ta ʃərət-we yɨ-wəra-n. be.hungry.PFV-3smS-3smO-CV.M-LINK food-DEF 3smS-eat.IPFV-3smO-M ‘Having become hungry he eats the food.’

This fact suggests that -we is likely to occur when a noun that should be definite (for whatever reason) has not been mentioned previously and cannot be re- ferred to anaphorically with the definite article -xʷɨt etc. This hypothesis seems to be supported by the situation found in the short excerpt in (815) where the speaker tells what he has heard about the disadvantages of early marriage and

²¹¹The definite article should not be confused with the we ‘or’ in questions or the (optional) question marker we in yes/no-questions.

236 giving birth. Since it is part of common knowledge that children belong to such a context and therefore are ‘known’, the first mention of tɨkə ‘child’ in the third sentence is with the definite article. However, due to the fact that it has not been mentioned explicitly in the previous sentences, anaphoric reference in the sense of “the aforementioned child” is not possible.

(815) bə-tkɨnət agβ-ot wəxe kʼar ã-xər-ə… yalə LOC-childhood marry-INF good THING NEG-become.PFV-3smS withoutѦ gɨzyə-xnəma cʼənə-ma-n-ta yɨ-tgʷədd-əma… tɨkə-we wəxe time-3pfPOSS give.birth.PFV-3pfS-CV.M-LINK 3S-be.hurt.IPFV-pfS child-DEF good kʼar e-trəkʼ… təxankʼʸətə tə-ms-əxnəma THING NEG.3smS-grow.up.IPFV afterwards COM-husband-3pfPOSS yɨ-trakəs-əma. 3S-quarrel.IPFV-pfS ‘Getting married in childhood is not good… giving birth without their period, they () are harmed… the child does not grow up well… and then they quarrel with their husband.’

Anyhow, these interpretations of -we are only tentative and have to be verified. Given that its frequency is low, it seems plausible that -we is indeed just a loan that does not play an integral role in the Gumer system of definiteness.

4.4.2 Use of articles and marking of definiteness Overt marking of definiteness in Gumer follows other rules than, for example, in European languages. The two main differences are, firstly, the rather clear distinction between direct anaphora and associative anaphora, and secondly the fact that the articles, in particular for direct anaphora, are much less obligatory and consequently less frequent.

4.4.2.1 Direct anaphora vs. associative anaphora In the marking of definiteness, Gumer distinguishes rather strictly between direct anaphora and associative anaphora.²¹² Cutting out the odd article -we, the definite articles xʷɨt(a) etc. as described in section 4.4.1 are used for direct anaphora (816), while associative anaphoric definiteness (817) is marked by the 3sm possessive suffix -əta (↗ 4.3.6).

(816) ema-xʷɨta ə-xʸɨr. way-DEF.sm 1sS-know.IPFV ‘I know the way (speaking about a known/previously mentioned way).’

²¹²Direct anaphora relates to a referent that has already been introduced to the discourse earlier (“the aforementioned”) and associative anaphora describes an indirect reference, where a referent is associated with another referent that has just been mentioned or to (extralinguistic) referents that are available in the discourse context (“the one related to the aforementioned”) (cf. Fraurud 2001).

237 (817) ema-ta ə-xʸɨr. way-3smPOSS 1sS-know.IPFV ‘I know the way (for example to the known place we intend to go).’

Example (818) illustrates the two different definiteness markings. First, a referent (kʷɨtara ‘chicken’) is introduced, which remains unmarked. A few sentences later the speaker resumes this same referent and marks it with the definite article -xʷɨta indicating that it is the known aforementioned chicken. In the following, a completely new referent (ɨxa ‘water’) enters the discourse. The topic of the story is cooking chicken and it is known to both speaker and hearer that chickens are cooked in water. Therefore, the first mentioning of ɨxa has to be marked as definite. Since it is not explicitly aforementioned but rather associated with kʷɨtara, the possessive -əta is in order.

(818) at kərə kʷɨtara (Ø) yɨ-cəkʷr-i […] kʷɨtara-xʷɨta cəkkər-xʷ-ɨn-ɨm. one day chicken 3S-cook.IPFV.IPS-3smO chicken-DEF.sm cook.PFV-1sS-3smO-M bə-cəkkər-xʷ-ɨn ankʼʸə ɨxa-ta xʷə-xʷ-ɨn-ɨm. LOC/TEMP-cook.PFV-1sS-3smO after water-3smPOSS spill.PFV-1sS-3smO-M ‘One day a chicken was cooked. […] I cooked the chicken. After I had cooked it, I spilled the water.’

As for the use of the definite article for direct anaphora, it is important to note that it is not used whenever an aforementioned referent occurs a second time. Rather it seems that a certain topicality is also needed for it to receive the definite article. In the second sentence of (819) fərəz ‘horse’ appears for the second time, however due to the fact that here it is a rather generic (backgrounded, almost “incorporated”) object of nəkʸəm ‘ride’ this happens without definite article. In contrast, it is the (new) topic and subject of the following sentence and marked definite.

(819) a. bə-raxʷ-i ankʼʸə, fərəz (Ø) tʼəβətʼ-ə-m LOC/TEMP-send.PFV.IPS-3smO after horse take.PFV-3smS-CV.M ankʼa-ta agəd-ə-m. […] mouth-3smPOSS tie.PFV-3smS-M ‘After they had sent him, he took a horse and tied its/the mouth.’ b. bet-ɨnyə tɨ-n-kəra-nə fərəz (Ø) yə-sost house-DIR TEMP-1pS-go.up.IPFV-1pS horse DAT-three nəkʸəm-ne-m. ride.PFV-1pS.3smO-M ‘When we went (up) home, the three of us mounted the horse.’ c. fərəz-xʷɨta t-i-yafəta nɨfas kʷəffɨyya horse-DEF.sm TEMP-3smS-gallop.IPFV wind hat wəsəd-ə-βi-m. take.PFV-3smS-MAL.1s-M ‘When the horse galloped, the wind took the hat from me.’

238 A more reliable description of the distribution definiteness markers requires further research with the aid of large text corpora, a task that clearly is beyond the scope of this thesis.

4.5 Numerals 4.5.1 Cardinal numerals

1 at 10 assɨr 100 bəkʼɨr~ məto 2 xʷet 20 xʷɨya 1000 xʷɨm~ ʃi 3 sost 30 sasa 4 arβət 40 arβa 5 ammɨst 50 amsa 6 sɨddɨst 60 sɨdsa ( ~sɨlsa) 7 səβat 70 sɨβa ( ~səβa) 8 sɨmʷɨt 80 sɨmra (~ səmana) 9 ʒətʼə 90 zɨtʼəra (~ zətʼəna)

Table 90: Cardinal numerals I

The numerals below ten and their corresponding multiples of ten share some sim- ilarities, but there is no regular derivative morpheme. Note that the final t of the numerals 2 to 8 (9 does not end in t) misses in the tens, which instead show a final a. As for ammɨst ‘five’, sɨddɨst ‘six’ and assɨr ‘ten’, they can also be heard without clear gemination. Probably due to school education, younger speakers occasionally also use the Amharic forms for higher numerals, for example sɨlsa instead of sɨdsa ‘sixty’ or zətʼəna instead of zɨtʼəra ‘ninety’. In the case of ‘hundred’ and ‘thousand’, the Amharic məto and ʃi (820) are much more in use than bəkʼɨr (821) and xʷɨm. In particular xʷɨm ‘thousand’ seems to become completely obsolete and is usually replaced by ʃi.

(820) at ʃi xʷet məto yə-mraxɨr kərə wə-βər-u? one thousand two hundred DAT-how.much day INF-say-COP.3smS ‘That is 1200 for how many days?’ (821) mʷena-ta togyə bəkʼɨr məgəra yə-sɛ. uncle-3smPOSS T. hundred calf 3smS-find.JUS ‘His uncles, the Togye, should find one hundred calves.’

239 counting quantity 11 asrat asrəm at 21 xʷɨyam at 12 asrə xʷet asrəm xʷet 22 xʷɨyam xʷet 13 asrə sost asrəm sost 47 arβam səβat 14 asrarβət asrəm arβət 99 zɨtʼəram ʒətʼə 15 asrammɨst asrəm ammɨst 117 məto asrə səβat 16 asrə sɨddɨst asrəm sɨddɨst 254 xʷet məto amsam arβət 17 asrə səβat asrəm səβat 838 sɨmʷɨt məto sasam səmʷɨt 18 asrə sɨmʷɨt asrəm sɨmʷɨt 1970 (at) ʃi ʒətʼə məto sɨβa 19 asrə ʒətʼə asrəm ʒətʼə 2606 xʷet ʃi sɨddɨst məto sɨddɨst

Table 91: Cardinal numerals II

The numerals from 11 to 19 are formed from asrə (rather than assɨr ‘ten’). There are two possibilities to compound them: either the unit numeral directly follows asrə, and in case the former begins with a vowel the ə of asrə is dropped; or there is an -m between asrə and the unit numeral (cf. -m ‘also, and’). It seems that the forms without -m are usually used for counting and the ones with -m to state a quantity. However, this distinction is probably not very clear-cut. Compound numerals above twenty tend to contain the -m both for counting and quantity (822)-(823), but there are also examples without it (824).

(822) b-ɨmmat kərə zətʼəna-m səβat səβ kʼəttʼər-o-m. LOC-only.one day ninety-ALSO seven person kill.PFV-3pmS-M ‘In a single day they killed 97 people.’ (823) xʷɨya-m arβət ɨkka banə-nə. twenty-ALSO four like.this BE.PT-1pS ‘We were like 24 [persons].’ (824) xʷɨya ammɨst bɨrr yɨ-bʷ-ɨn. twenty five birr 3S-give.IPFV.IPS-1sO ‘They give me 25 birr.’

To express totality (‘all’), another -m is suffixed to the cardinal numerals and the final t of 2 to 8 is palatalized to c. Thus we have, for example, xʷecɨm ‘all two, both’ (825), soscɨm ‘all three’ (826), arβəcɨm ‘all four’, ʒətʼəm ‘all nine’, amsam ‘all fifty’, mətom ‘all hundred’, ʃim ‘all thousand’. The final t of at ‘one’ in compound numerals is not palatalized, as for example in xʷɨyam atɨm ‘all twenty one’.

(825) yə-xʷecɨm-ənda at-at adda aw-ɨndə-m. DAT-both-1pPOSS one-one servant give.PFV.IPS-1pO-M ‘For both of us they gave one servant each.’ (826) soscɨm gɨzyə yɨ-srə-βo-nku banə? all.three time 3S-buy.IPFV-pmS-BEN.2pm AUX.PT ‘They bought (would buy) all three meals (lit. times) for you?’

240 4.5.2 Ordinal numerals

1st atənə 10th assɨrənə 100th bəkʼrənə 2nd xʷetənə 20th xʷɨyanə 1000th xʷɨmranə 3rd sostənə 30th sasanə 4th arβətənə 40th arβanə 5th ammɨstənə 50th amsanə 6th sɨddɨstənə 60th sɨdsanə 7th səβatənə 70th sɨβanə 8th sɨmʷɨtənə 80th sɨmranə 9th ʒətʼənə 90th zɨtʼəranə

Table 92: Ordinal numerals

Ordinal numerals are formed by -ənə (-nə after a vowel) suffixed to the cardinal numeral. Some people tend to employ the Amharic -əɲɲa instead of -ənə, especially for higher numerals. Note that the ordinal numeral of xʷɨm ‘thousand’ is xʷɨmranə. Like all modifiers the ordinal numerals precede the noun (827).

(827) bə-mɨr-u səβatənə fokʼ ə-wətʼa? INSTR-what-COP.3smS seventh floorѦ 1sS-go.upIPFV ‘With what do I go to the seventh floor?’

Finally, take note of the names of the fractions gɨβt ‘hal’ (828), kʼɨrtʼ ‘quarter’ and bʷɨtʼɨr ‘eighth’. The latter one is often used to designate a part of the wɨssa- bread that is not necessarily the eighth part of it. Finally, shares are counted by means of əj ‘hand’, for example sost əj ‘three shares’, ammɨst əj ‘five shares’, etc.

(828) yə-βora bəsər gɨβtɨ-gɨβt yə-sya-m-ənda afettən-ne-m. ATTR-ox meat half-half DAT-joke-ALSO-1pPOSS finish.up.PFV-1pS.3smO-M ‘We easily finished up the ox meat each a half.’

4.6 Days of the week The days of the week are shown in table 93. According to Leslau (1979b: 260) the name of Friday can also be addərə in Chaha and most other Gurage varieties, but I have only recorded jɨmat. Due to its Arabic origin (

241 Monday wɨtʼət Tuesday wɨtʼət mərəga Wednesday əro Thursday amʷɨs Friday jɨmat Saturday kʼətʼa səmbət Sunday wɨr səmbət

Table 93: Days of the week

4.7 Nominal affixes and subordinators Gumer possesses a relatively small set of affixes, both prefixes and suffixes, that occur with nominals. All affixes except the associative marker nə- and -nyə ‘towards’ also appear with verbs to form various kinds of subordinated clauses. Ta- ble 94 summarizes the affixes and their basic functions both with nominals and in subordination.

Nouns Subordinated verbs yə-  /   bə-/ b- ,  ,  tə-/ t-  /  , () nə-  — -e   -xəma  ,  -nyə  —

Table 94: Nominal affixes and subordinators

The three prefixes yə-, bə-, and tə-, very likely older than the suffixes, are mutually exclusive. As opposed to this, combinations of prefixes and suffixes as well as successions of two suffixes are possible. Both -xəma and -e often occur together with yə-, and -nyə is most of the time preceded by -e. The former case is reminiscent of the co-occurrence of prefixes and postpositions to express some more specialized relations (↗ 4.7.1.6). All affixes are in their essence nominal rather than verbal markers. The fact that they also appear with verbs can be explained – or at least visualized – as follows: for subordination only the so-called ‘relative verbs’ (↗ 4.7.3.1) can be used. The ‘relative verbs’ themselves are a ‘nominalized’ form of the verb that are used for relative clauses, i.e. to modify nominals, but also as (headless) nominals themselves comparable in their function but not in their form to participles in European languages. As ‘nominalized’ forms, the ‘relative verbs’ can be furnished with nominal morphology like articles on the one hand, or the said affixes to build

242 forms used in subordination expressing more or less specific meanings on the other hand. Nevertheless, it is very important to note here that the claim that the ‘relative verbs’ are ‘nominalizations’ of verbs is not at all obvious and therefore highly disputable. In principle, ‘relative verbs’ are formed by the attributivizer yə-, but this morpheme is only present, i.e. “surfacing”, in the Perfective and only when there is no other prefix involved. Thus, in many instances no visible nominalizing morpheme as such is present. The properties of ‘relative verbs’ are discussed in more detail in section 4.7.3.1. There are a few further suffixes (or clitics) that have quite a different distribution than the above affixes. They are not nominal markers per se but mainly serve pragmatic functions and/or information structure, like -ʃ, -x and -a.²¹³ The suffix -m represents an intermediate case. It is used as pragmatic and focusing marker (‘also’) (829), as nominal coordinator (‘and’) (830), and as linker with verbs (not based on ‘relative verbs’) forming the so-called m-converbs (↗ 3.14.2).

(829) zətʼəna ammɨst bɨrr bet-ɨm y-asəra-xu? ninety five birr house-ALSO 3smS-bring.IPFV-2pmO ʽDo 95 birr bring you also home?ʼ (830) gərəd-ɨm ərc-ɨm girl-ALSO boy-ALSO ʽa girls and a boyʼ

4.7.1 Affixes and nouns: prefixes The most versatile prefix in Gumer is yə- which combines a variety of functions. Firstly, it is an attributivizer that conjoins (two or more) nouns. Secondly, it covers dative-like functions marking overt nominals as recipients, beneficiaries (and maleficiaries), and by extension in some cases also primary objects (i.e. differential object marking); and furthermore it is also used to mark peripheral constituents that are not local, instrumental or comitative, often translating with ‘for’. The latter functions are covered by the prefixes bə- ( and ) and tə- ( and ). The attributivizer yə- is dropped when bə- or tə- is present. The same is true with the associative marker nə-, which is described in section 4.2.2.

4.7.1.1 Attributivizer yə- A noun is attributed to another noun with yə- on the first element – the dependent noun or ‘possessor’ – followed by the head noun which remains unmarked.

(831) yə-kəbbədə ɨmɨr ATTR-K. stone ‘Kebbede’s stone’

²¹³Very roughly speaking, -ʃ corresponds to ‘you know’, -x to ‘as for’, and -a to ‘I assure you (em- phasis)’.

243 There are basically two possibilities how the attributed noun qualifies the head noun. First, the common relation is possessor–possessed in very broad sense ranging from ownership to a rather general notion of ‘belonging’ that expresses that an entity relates to another one.

(832) yə-βeyda xʷɨjɨr ATTR-B. clothes ‘Beyda’s clothes’ (833) yə-jəmal adot ATTR-J. mother ‘Jemal’s mother’ (834) yə-dina gən ATTR-outsider country ‘the enemy’s country’ (835) yə-gʷrage kʼar ATTR-Gurage language ‘the language of the Gurage’ (836) yə-gʷeta fəkʼad ATTR-God permissionѦ ‘God’s permission’ (837) yə-jəβən ɨffʷət ATTR-coffee.pot pot.lid ‘the lid of the coffee pot’ (838) yə-drə səβ ATTR-former.times person ‘people of/in old times’

Second, the attributed noun can modify the meaning of the head noun specifying its type, subcategory, origin, material, etc.

(839) yə-ʒɨr danə ATTR-zhir judge ‘zhir-judge’²¹⁴ (840) yə-wərkʼ kʼələβət ATTR-gold ring ‘gold ring’

The distinction between possession/belonging and specification can be very subtle. Consider the different possible translations of the following examples.

²¹⁴ʒɨr is a unit to measure land (approximately 12 feet) (Leslau 1979a: 136) and the zhir-judge is the person in charge of measuring the land.

244 (841) y-asa ʃərət ATTR-fish food ‘food [made] of fish’ / ‘fish food (i.e. food for fish)’ (842) y-əram goga ATTR-cow skin ‘cow hide (material)’/ ‘the cow’s skin (possession)’

Further, note that instead of the ordinary possessive suffixes (↗ 4.3.6) one can also attribute the full pronouns to express possession by means of yə-.

(843) y-ɨyya angət ATTR-1s neck ‘my neck’ (844) y-axə ʃɨm ATTR-2sm name ‘your (sm) name’

Nouns with yə- can also occur elliptically without head noun as in (845)-(846).

(845) y-arɨβ sera-w-ʃ, yə-tʼorɨnnət. ATTR-war custom-COP.3smS-PRAG ATTR-warѦ ‘That is the customs of war, of fighting.’ (846) A: nam-e ɨnde sɨlk-əxʸta nɨ-məzgɨbʷ-ɨn. give.IMP-2smS-1sO please phoneѦ-3sfPOSS 1sS-registerѦ.JUS-3smO ‘Give [it to] me please, so I can register her phone [number].’

B: əgi, yə-rəwda we yə-jəmal? okay ATTR-R. or ATTR-J. ‘Ok, Rauda’s or Jemal’s?’

When there is more than one noun attributed, the marker yə- occurs only on the first element.

(847) yə-βora bəsər kɨtfʷə (*yə-βora yə-βəsər kɨtfʷə) ATTR-ox meat kitfo ‘kitfo [made] of ox meat, ox meat kitfo’ (848) yə-soresa bet xʷɨjɨr yə-βora dannəra-βa. (*yə-soresa yə-βet xʷɨjɨr) ATTR-hero house clothes ATTR-ox tanned.hide-AUX.PT ‘The clothes of a hero’s house (family) were (o) ox leather.’ (849) y-adot-ənda abba bet (*y-adot-ənda y-abba bet) ATTR-mother-1pPOSS father house ‘the house of the father of our mother’

The attributivizer yə- disappears when one of the two prepositions bə- or tə- is attached.

245 (850) b-əcʼə kʼʸɨn-e < *bə-yə-əcʼə kʼʸɨn-e LOC-wood bottom-GOAL ‘at the bottom of the tree’ (851) bə-sost səβ gɨβt < *bə-yə-sost səβ gɨβt LOC-three person middle ‘in the middle of three persons’ (852) b-atʼatʼ ema acənə-xu-m-tanə tə-drɨjjɨt ema LOC-A. road bring.PFV-2pmS-CV.M-LINK COM-organizationѦ road dəməd-xu-m. < *bə-y-atʼatʼ; < *tə-yə-dɨrɨjjɨt join.PFV-2pmS-M ‘You brought [it] to the road of Atʼatʼ and connected [it] with the road of the organization.’

4.7.1.2 Recipients, beneficiaries/maleficiaries, primary objects marked by yə- The second major function of the prefix yə- is the marking of dative-like roles, namely recipients (853) including addressees (854), further beneficiaries (855) and maleficiaries (856), and finally certain direct objects (857).

(853) sat-əta yə-gərəd aβ-ə-m. watch-3smPOSS DAT-girl give.PFV-3smS-M ‘He gave his watch to a girl.’ (854) y-asən bɨrr name bɨr-ot kʼəmətʼ-u. DAT-A. birr give.IMP[.2smS].1sO say-INF shame-COP.3smS ‘It is a shame to say to Hasen: “Give me money”.’ (855) tʼay-xʷɨta y-ərc-əta sɨyə-lə-m. sheep-DEF.sm BEN-son-3sm.POSS buy.PFV[.3smS]-BEN.3sm-M ‘He bought the sheep for his son.’ (856) əkkʷa yə-kəbbədə gʷəncə cənə-wə-m. today DAT-K. hyena come.PFV[.3smS]-MAL.3sm-M ‘Today a hyena came to Kebbede (which is dangerous).’ (857) yə-kəbbədə kʼʷəttʼər-ə-n-ɨm. DAT-K. kill.PFV-3smS-3smO-M ‘He killed Kebbede.’

The marking of some direct objects - known as differential object marking - is conditioned by an interplay of several factors. Overt nouns (or noun phrases) are more likely to receive yə- the higher they are on the hierarchies of definiteness, humanness, animacy, topicality or discourse prominence, but also to disambiguate when a referent could be understood either as subject or object.

4.7.1.3 Peripheral constituents marked by yə- As mentioned above, yə- is also used to mark several kinds of peripheral relations that are not local, instrumental or comitative. Many uses correspond to English

246 ‘for’ as in examples (858)-(863) below. Consider also the question words yə-nkʼar (< yə-mɨkʼar) and yə-mɨr both meaning ‘for what, why’.

(858) yə-zər wənd-ɨxə-m-ba emra? DAT-rainy.season go.down.PFV-2sm-M-AUX.PT last.year ‘Did you go down (to the countryside) for the rainy season last year?’ (859) at ʃi xʷet məto yə-mraxɨr kərə wə-βər-u? one thousand two hundred DAT-how.much day INF-say-COP.3smS ‘That is 1200 for how many days?’ (860) yə-transɨport kas-o-ndə-m. DAT-transportation pay.PFV-3pmS-1pO-M ‘They paid us for transportation.’ (861) yə-xʷya at wəfer DAT-twenty one young.bull ‘for twenty [calves] one young bull’ (862) bazəra y-at kərə yə-xʷet kərə yə-sost kərə təmettər-ə-m guest DAT-one day DAT-two day DAT-three day be.stretched.PFV-3smS-CV.M yɨ-tkʼʸəppɨr. 3smS-accept.IPFV ‘A guest (stranger) is accepted for one day, for two days, for three days maximum.’ (863) yə-x bə-tət yə-gʸəta yə-xɨr. DAT-DEM LOC-down DAT-Gyeto 3smS-become.JUS ‘From there down it (the land) shall be for the Gyeto.’

Further occurrences of yə- cover many different uses translated into English as ‘about’, ‘by’, ‘in’, etc., as illustrated in the following examples (864)-(869). Note also the forming of the adverbial expression in (870).

(864) əxʷa yə-xʷt kʼar yə-tcʼawəj-i. now DAT-3sm THING 3S-talk.JUS.IPS-3smO ‘Now let’s talk about that.’ (865) yə-gʷəmarə kʼar ə-tʼəf-te. DAT-Gumer language 1sS-write.IPFV-FUT.DEF ‘I will write about the Gumer language.’ (866) yə-xʷjɨr-əxno tʼatʼa enə-no. DAT-clothes-3pmPOSS troubleѦ NEG.EX[.3smS]-3pmO ‘They do not have any trouble with/about their clothes.’ (867) at ʃi xʷet məto yə-xʷya t-i-ʃəj-i one thousand two hundred DAT-twenty TEMP-3S-divide.IPFV.IPS-3smO mɨraxr-u? how.much-COP.3smS ‘How much is it when you divide 1200 by twenty?’

247 (868) yə-gʷragina zəngʸ! DAT-Gurage.language speak.IMP[.2smS] ‘Speak (in) Gurage!’ (869) fərəz yə-sost nəkʸəm-nə-m. horse DAT-three ride.PFV-1pS-M ‘We mounted the horse in three (at a time).’ (870) a. yə-sya-m-əta DAT-joke-ALSO-3smPOSS ‘easily, simply, not seriously’ b. yaβekʼo lallego yəsyaməta yəsyaməta Y. L. easily easily sədəd-ə-n-ɨm-ʃ. drive.away.PFV-3smS-3smO-M-PRAG ‘Yabeko Lallego drove him away very easily.’

4.7.1.4 Locative bə- and instrumental bə- The preposition bə- covers two different basic functions. Firstly, it expresses locative meanings (: ‘in, at’) as illustrated in (871)-(874). This includes also temporal meanings (875)-(877).

(871) bə-gən-əxʷna nɨkkʼar yɨ-kʷəʃ-i, an-xər-ə-we? LOC-country-3pmPOSS much 3S-pay.IPFV.IPS-3smO NEG-become.PFV-3smS-Q ‘In their country they pay a lot, don’t they?’ (872) bə-joka attər-o-m. LOC-J. spend.the.night.PFV-3pmS-M ‘They spent the night in Joka.’ (873) bə-fraʃ attən-nə-m. LOC-mattress spend.the.night.PFV-1pS-M ‘We spent the night on mattresses.’ (874) bə-z mədər kʼʷəyə-nǃ LOC-DEM place wait.IMP[.2smS]-3smO ‘Wait for him hereǃ’ (875) b-ɨmmat kərə zətʼənamsəβat səβ kʼəttʼər-o-m. LOC-only.one day ninety.seven person kill.PFV-3pmS-M ‘They killed ninety-seven persons in a single day.’ (876) ɨnnɨm gɨzyə bə-smʷɨt sat yɨ-cən. every time LOC-eight hour 3smS-come.IPFV ‘He always comes at eight oʼclock.’ (877) bə-tkɨnət agβ-ot wəxe kʼar ã-xər-ə. LOC-childhood marry-INF good THING NEG-become.PFV-3smS ‘Getting married in childhood is not good.’

Secondly, bə- marks instrumental meanings (: ‘with, by means o’) as shown in examples (878)-(880).

248 (878) b-ɨxa-xɨʷta ɨngʷəd mena tɨ-cot-wə. INSTR-water-DEF other work 3sfS-make.IPFV-MAL.3sm ‘She makes something else with the water.’ (879) b-alanga²¹⁵ t-i-dərg-e gɨzyə INSTR-whip TEMP-3smS-hit.IPFV-1sO time ‘when he hit me with a whip’ (880) bə-kʼɨβ əsəssəw-i-m. INSTR-butter massage.PFV.IPS-3smO-M ‘One was massaged with butter.’

4.7.1.5 Comitative tə- and ablative tə- The preposition tə- has two different basic functions. On the one hand, it expresses a comitative meaning (: ‘with, together with’) as in (881)-(884).²¹⁶ By extension, tə- is also used as coordinating device that usually translates as ‘and’. While the neutral coordination is formed with …-m …-m (more literally ‘also … also …ʼ), tə- occurs in coordination of two nouns that belong together closely or even naturally and appear often in collocation. Consider for instance the word for ‘parents’ aβtadot < aβ t-adot (i.e. father -mother), further examples being the proper names of a pair of siblings (886) or ‘Chaha and Gumer’ (885), as well as the two adjectives referring to two brothers (887).

(881) tə-gʷəppɛ-yəna sost fərəz tʼəβətʼ-nə-m wəttʼa-nə-m. COM-brother-1sPOSS three horse take.PFV-1pS-M go.up.PFV-1pS-M ‘Together with my brothers we took three horses und went up.’ (882) tə-ms-əxnəma yɨ-trakəs-əma. COM-husband-3pfPOSS 3S-quarrel.IPFV-pfS ‘They quarrel with their husbands.’ (883) t-aβdul mənnan tərakəβ-nə-m. COM-A. M. meet.PFV-1pS-M ‘We met Abdul Mennan.’ (884) t-axə tor-ot ə-ʃə. COM-2sm sit-INF 1sS-want.IPFV ‘I want to sit with (next to) you.’ (885) cəxa tə-gʷəmarə ɨmmat-u əgr-angət cəxa yɨ-wr-i. Chaha COM-Gumer only.one-COP.3smS leg-neck Chaha 3S-say.IPFV.IPS-3smO ‘Chaha and Gumer are the same, they are called egranget Chaha.’ (886) sidi ɨmɨryə tə-gəβrə_xanna cʼənə-m. S. I. COM-G._H. give.birth.PFV[.3smS]-M ‘Sidi fathered Imirye and Gebre Hanna.’

²¹⁵Leslau (1979c: 40) has aranga for Chaha and alanga for Inor. ²¹⁶Note the different constructions of Gumer and English in (883), the English verb ‘meet’ not re- quiring the preposition ‘with’.

249 (887) ɨrs tə-rkʼ gʷəppay bə-xər-o little COM-big brother COND-become.PFV-3pmS ‘if they are a younger and an older brother’

On the other hand, tə- expresses ablative meanings (: ‘from’) as in examples (888)-(890). This also includes comparative constructions where tə- marks the standard of comparision (‘than’), shown in examples (891)-(892).

(888) tə-mena gəppa-c-ɨm ʃərət tɨ-t-cot tɨ-tnadəd. ABL-work enter.PFV-3sfS-CV.M food TEMP-3sfS-work.IPFV 3sfS-be.angryѦ.IPFV ‘When she comes home from work and prepares food, she gets angry.’ (889) yə-gʷrage y-atʼəfɛ ɨras gobəna tə-tət cənə-m... DAT-Gurage 3smS-ambush.IPFV:PURP R. G. ABL-down come.PFV[.3smS]-M tə-cəxa cənə-m. ABL-Chaha come.PFV[.3smS]-M ‘To ambush the Gurage, Ras Gobena came from below... he came from Chaha.’ (890) wərəjə-ta xʷet əcʼɨr-u, tə-zɨkkɨm at-u, space.in.front.of.house-3smPOSS two fence-COP.3smS ABL-like.this one-COP.3smS tə-xɨkkɨm at-u. ABL-like.that one-COP.3smS ‘The space in front of the house is [defined by] two fences, one from here, one from there.’ (891) bet-əna tə-βet-axə yɨ-rkʼ. house-1sPOSS ABL-house-2smPOSS 3smS-be.bigger.IPFV ‘My house is bigger than your house.’ (892) tə-βɨrtɨkʷan lomi yɨ-fəzz. ABL-orange lemonѦ 3smS-be.better.IPFV ‘Lemons are better than oranges.’

4.7.1.6 Combinations with postpositions (relational nouns) In addition to the affixes there are several postpositions that either occur alone or, more commonly, together with one of the three relational prefixes (↗ 4.7.1). Some of them correspond to nouns designating body parts as indicated in the not exhaustive list in (893). Some (local) postpositions often feature an additional suffix -e (↗ 4.7.2); while yɨfte ʽbeforeʼ only exists with -e, there is, for instance, a variation ankʼʸə~ ankʼʸe ʽafterʼ.

(893) ankʼʸə ʽafter, behindʼ < back²¹⁷ dar ʽuntilʼ < border, limit dən ʽinsideʼ < belly eβar ʽwithoutʼ fʷər ʽabove, onʼ kʼʸɨn ʽ(at) bottomʼ < bottom, buttocks

250 məyə ʽbeside, alongside, next toʼ < side, rib sɨn ʽuntilʼ < səna √srA ʽreachʼ tət ʽbelow, underʼ yɨft-e ʽbefore, in frontʼ < face- yɨft-ɨft ʽopposite, in front ofʼ < face-face

The following examples illustrate the uses of some postpositions without prefixes (894)-(896) and with prefixes (897)-(900). See also the formation of ʽafterwardsʼ in figure 8.

(894) əxʷa-m dar yɨ-rəβɨr. now-ALSO until 3smS-live.IPFV ‘He lives until now (i.e. still).’ (895) ɨzəxʷəce y-agaz andenə kʼəβɨr sɨn ɨnnɨm y-adya banə. I. ATTR-A. A. grave until all ATTR-Siltʼe BE.PT ʽIzekhweche until the grave of Agaz Andene all belonged to the Siltʼe.ʼ (896) ʃay sɨkkʷar eβar a-n-ʃə. tea sugar without NEG-1sS-want.IPFV ‘I do not want tea without sugar.’ (897) wəʃər-xʷɨta bə-βet yɨftɨft awəna-n-ɨm. cooking.pot-DEF.sm LOC-house opposite put.PFV[.3smS]-3smO-M ‘He put the cooking-pot opposite (in front o) the house.’ (898) b-əcʼə fʷər wədərə səkkʼər-ə-m. LOC-wood top rope hang.PFV-3smS-M ‘He hung a rope on a/the tree (lit. wood).’ (899) tə-βet fʷər wənd-ə-m. ABL-house top descend.PFV-3smS-M ‘He came down from (on top o) the roof.’ (900) yə-βr-ot-ənda yɨfte dəwwəl-xʷ-ɨlə-m. DAT/ATTR-eat-INF-1pPOSS before phoneѦ.PFV-1sS-BEN.3sm-M ‘Before we ate I called him.’

4.7.2 Affixes and nouns: suffixes The suffixed similative -xəma ʽlike, ʼ (901)-(902) and the directional(-local) -e ʽto, (at), ʼ (903) occur together with the prefix yə- (here glossed as  in analogy to other peripheral constituents, ↗ 4.7.1.3).

(901) yə-wəndɨmu-xəma DAT-W.-SIMIL ʽlike Wendimuʼ

²¹⁷The word ankʼʸə does not denote ‘back’ as bodypart in Gumer (or Chaha), but take notice that it has this meaning in Mesqan and Siltʼe (Leslau 1979c: 72).

251 (902) y-axə-xəma ɨmbi-n-xʷ. DAT-2sm-like refusalѦ-COP-1sS ʽI am undefeated like you.ʼ (903) sost fərəz tʼəβətʼ-nə-m wəttʼa-nə-m, yə-cʼɨʃt-e. three horse take.PFV-1pS-CV.M go.up.PFV-1pS-M DAT-Ch.-GOAL ʽWe took three horses and went up, to Chisht.ʼ

However, -e occurs without yə- with demonstratives forming spatial adverbs (↗ 4.3.3.2).

(904) zɨx-e tɨ-tmər-e cənə-c-ɨm. DEM-GOAL 3sfS-learn.IPFV-PURP come.PFV-3sfS-M ʽShe came here to study.ʼ

In combination with bə-, -e expresses a local meaning rather than a direction or goal. Note again that yə- is dropped when another preposition is present.

(905) bə-βxʸ-e tərakəβ-nə-m LOC-funeral-GOAL meet.PFV-1pS-M ʽWe met at the funeral.ʼ

The directional -nyə translates roughly as ʽtowards, ʼ and partly overlaps with -e semantically (906)-(907). Often it is preceded by -e as in (908)-(909), reinforcing the directional meaning of -e.

(906) gən-ɨnyə an-tɨʒəppər-xə? countryside-DIR NEG-return.PFV-2smS ‘You did not return to the countryside?’ (907) bet-ɨnyə ə-tgəttər-e wər-xʷ-ɨm. house-DIR 1sS-lie.down.IPFV-PURP go.PFV-1sS-M ‘I went home to sleep.’ (908) ɨngʷəd-e-nyə a-n-ar. other-GOAL-DIR NEG-1sS-go.IPFV ʽI do not go anywhere else.ʼ (909) xɨx-e-nyə tʼən-o-ndə-m. DEM-GOAL-DIR call.PFV-3pmS-1pO-M ʽThey called us there.ʼ

4.7.3 Affixes and verbs: subordinate clauses 4.7.3.1 ʽRelative verbʼ As introduced in section 4.7, the traditionally called ʽrelative verbsʼ can be con- ceptualized as nominalized verb forms. Firstly, they work like adjectives, i.e. they occur before nouns and modify them as in (910)-(911), in other words they form relative clauses (hence the name ʽrelative verbʼ). Secondly, they can host nominal morphology such as the definite article (912)-(914) or possessive suffix (915).

252 (910) yə-mʷət-ə abba-na REL-die.PFV-3smS father-1sPOSS ʽmy father who diedʼ (911) yɨ-cən amət [REL.]3sm-come.IPFV year ʽnext/coming year (lit. year that comes)ʼ (912) yə-dənəs-ə-n-xʷɨta REL-sing.and.dance.PFV-3smS-3smO-DEF.sm ʽ(that) what he sangʼ (913) y-oj-yo-xɨno REL-tell.IPS.PFV-3pmO-DEF.pm ‘the ones who were told’ (914) y-axə yɨ-kəs-xə-xʷɨt DAT-2sm [REL.]3smS-pay.IPFV-2smO-DEF.sm ʽthe one who pays youʼ (915) sər yɨ-fʷacʼ-əta sər yɨ-fʷacʼ. grass [REL.]3smS-mow.IPFV-3smPOSS grass 3smS-mow.IPFV ʽThe one who mows (i.e. is able to mow) grass, mows grass.ʼ

The ʽrelative verbʼ can be  or  (but not ). The  marker is the prefix yə-, i.e. the same as the attributivizer (↗ 4.7.1.1), but the catch in Gumer is that yə- only appears with , while with  there is no overt marker. Moreover, negated  often lack yə- as well. Nevertheless,  ʽrelative verbsʼ are diagnosed by the properties mentioned above and illustrated in (911), (914) and (915): their position before the head noun and the possibility of hosting nominal morphology. The primary use of the ʽrelative verbsʼ is in relative clauses (↗ 4.7.3.2). Fur- ther, as headless ʽpseudo-relativesʼ, i.e. not attributed to a noun, they express the backgrounded unfocused old information in cleft sentences, here illustrated with focused question words (916)-(917). Again, note that the ʽrelativeʼ  is not distinguishable from a main verb  formally, but functionally by its occurence in the cleft construction (cf. with the unclefted counterparts).

(916) mɨkʼar-u yə-tənəf-ə? (cf. mɨkʼar tənəf-ə-m) what-COP.3smS REL-remain.PFV-3smS what remain.PFV-3smS-M ʽWhat is left? (lit. what is it that remained)ʼ (917) yənkʼar-u yɨ-dəkʼ? (cf. yənkʼar yɨ-dəkʼ) why-COP.3smS [REL.]3smS-laugh.IPFV why 3smS-laugh.IPFV ‘Why is he laughing?’

Finally, the ʽrelative verbsʼ can be viewed as the basis of any other subordinate clause involving the affixes b(ə)-, t(ə)-, -e, -xəma (see table 94) as shortly described in the sections 4.7.3.3-4.7.3.7 below. The marker of the ʽrelative verbsʼ yə- is indicated in the glosses with  only when overtly present. However, in this section (4.7.3) all ʽrelative verbsʼ are glossed with [.] for illustrative purposes.

253 4.7.3.2 Relative clauses To form relative clauses the ʽrelative verbʼ (↗ 4.7.3.1) is employed and constitutes (as in any other subordinate clause) the last word of the clause. Thus it directly stands before the head noun it modifies, with all other constituents of the relative clause preceding the ʽrelative verbʼ. When the head noun is the subject of the relative clause, the verb agrees regularly with it as in (918)-(919).

(918) bə-tʼərəppʼeza fʷər yə-cona səβ LOC-tableѦ on REL-sit.PFV[.3smS] person ʽ(a/the) person who sits on (top o) the tableʼ (919) kabbort yə-txəttər-əma ɨʃta coatѦ REL-dress.PFV-3pfS women ʽwomen wearing (lit. who have put on) coatsʼ

In the other cases, the head noun is resumed pronominally in the relative clause according to its syntactic role. If it is the primary object (920),  (921) or  (922), it is marked with the corresponding suffix on the ʽrelative verbʼ.

(920) gʷəyə yɨ-wr-i mədər G. [REL.]3S-say.IPS.IPFV-3smO place ʽ(a/the) place called Gweye (lit. a place that one calls [it] Gweye)ʼ (921) basta yə-ʃəkət-əc-lo gəmya gʷəppay-əxta-l-o. pasta REL-prepare.PFV-3sfS-BEN.3pm men brother-3sfPOSS-COP-3pmS ʽThe men she prepared pasta for are her brothers (lit. that she prepared pasta for them).ʼ (922) t-adr-o-pʷə mədər [REL.]2S-spend.the.night.IPFV-pmS-MAL.3sm place ʽ(the) place where you (pm) spend the night (lit. that you spend the night in it)ʼ

4.7.3.3 Temporal clauses The basic temporal clauses (ʽwhenʼ) are formed with t- +  (923) or b- +  (924).

(923) dengʸa-ta t-i-ʃə amədar kʼʷəttʼər-ə-n-ɨm. children-3smPOSS TEMP-3smS-want.IPFV coldness kill.PFV-3smS-3smO-M ʽWhen he was looking for his children, coldness killed him.ʼ (924) mɨr gən-l-o b-i-βɨr gʷəmarə-l-o what country-COP-3pmS TEMP-3smS-say.IPFV Gumer-COP-3pmS bʷar-i-m. say.PFV.IPS-3smO-M ʽWhen he said “What country are they [from]?”, one said “They are [from] Gumer”.ʼ

254 Both options are often interchangeable and moreover can overlap with conditional meanings (↗ 4.7.3.5). However, explicitly simultaneous events (ʽwhileʼ) tend to be expressed with t- . Furthermore, the repetition of verbs with t-  expresses long-lasting events (925).

(925) tɨ-n-afəta tɨ-n-afəta bet cənə-xʷ-ɨm. TEMP-1sS-ride.IPFV TEMP-1sS-ride.IPFV house come.PFV-1sS-M ‘Riding riding, I came home.’

Temporal clauses are often supplemented with postposed gamʷə ~gɨzyə ~gi ʽtimeʼ (926)-(928).

(926) sɨddət t-iy-ar-o gamʷə at gənə wər-o-m exile TEMP-3S-go.IPFV-pmS time one country go.PFV-3pmS-CV.M attər-əβo-m. spend.the.night.PFV-3pmS-M ʽWhen they went into exile, they went to one (a certain) country and spent the night.ʼ (927) bə-mʷət-ə gɨzyə zɨx ərc səβɨr bəkkər-ə-m. TEMP-die.PFV-3smS time DEM boy patience lack.PFV-3smS-M ‘When he died, this boy lost (his) patience.’ (928) a-n-ar t-i-βɨn-no gi... NEG-1sS-go.IPFV TEMP-3smS-say.IPFV-3pmO time ʽWhen he said to them “I donʼt go”...ʼ

Negated temporal clauses translate as ‘before’ (929)-(930) or ‘without’ (931)- (932) (also see example (477b)). Note that most of the time they feature an additional focusing -m (ʽʼ). The same form is also used to negate converbs as is the case in the latter two examples.

(929) cʼet t-e-wətʼa əray ərəβ-ot yɨ-kəβd. sun TEMP-NEG.3smS-ascend.IPFV cows milk-INF 3smS-be.difficult.IPFV ʽIt is difficult to milk cows before the sun rises.ʼ (930) gəβya t-e-sər-o-m kənə-nyə ʒor-o-m. market TEMP-NEG.3S-arrive.IPFV-pmS-ALSO right-DIR turn.PFV-3pmS-M ‘They turned to the right before they reached the market.ʼ (931) t-a-m-bəra-nə-m wan-nə-m. TEMP-NEG-1pS-eat.IPFV-1pS-ALSO spend.the.day-1pS-M ʽWe spent the day without eating (i.e. we didnʼt eat the whole day).ʼ (932) t-e-xʸr-ɨm dənəgʷ-ə-n-ɨm. TEMP-NEG.3smS-know.IPFV-ALSO hit.PFV-3smS-3smO-M ʽHe beat him without knowing (i.e. unintentionally).ʼ

The temporal relation ʽafterʼ is not expressed with an  temporal verb but by bə- (or tə-) plus (relative)  verb and the postposition ankʼʸə~ ankʼʸe ʽafterʼ (933)- (935), analogous to nominal local/temporal relations with postpositions (↗ 4.7.1.6).

255 (933) bə-jəppər-əc ankʼʸə LOC/TEMP-[REL.]finish.PFV-3smS after ʽafter she had finished (after finishing)ʼ (934) təxankʼʸe bə-cənə-xə ankʼʸə nɨβrət bəmɨr-βa? afterwards LOC/TEMP-[REL.]come.PFV-2smS after life how-BE.PT ʽAnd then, after you had come, how was life?ʼ (935) əj-əna t-antʼ-ɨxʷ ankʼʸe accɨm an-fʷaccʼə-xʷ. hand-1sPOSS ABL/TEMP-[REL.]cut.PFV-1sS after at.all NEG-mow.PFV-1sS-M ʽAfter I (had) cut my hand, I never mowed again.ʼ

Further, -xəma ‘like, ’ suffixed to bə-/tə- + (relative)  expresses ‘just/exactly when, immediatelyʼ (936).

(936) bet tə-gəppa-xʷ-xəma əram cʼənə-m. house ABL/TEMP-[REL.]enter.PFV-1sS-SIMIL cow give.birth.PFV[.3smS]-M ‘Exactly when I entered the house, a cow gave birth.’ (937) bəze eβ t-anəw-i-m-xəma yɨ-səcʼ-i. here milk ABL/TEMP-[REL.]milk.PFV.IPS-3smO-ALSO-SIMIL 3S-drink.IPFV.IPS-3smO ‘Here milk is drunk immediately after milking.’

4.7.3.4 Complement clauses Complement clauses are formed by -xəma suffixed to the ʽrelative verbʼ, which can be  (938)-(939) or  (940)-(941) depending on the intended tense/aspect of the subordinate clause.

(938) bete yə-trakəβ-nə-xəma tərəss-e-m. where REL-meet.PFV-1pS-COMP forget.PFV-3smS.1sO-M ʽI (have) forgot where we met.ʼ (939) abba-ta yə-mʷət-ə-xəma y-ud-xə-te. father-3smPOSS REL-die.PFV-3smS-COMP 3smS-tell.IPFV-2smO-FUT.DEF ‘He will tell you that his father died.’ (940) abba-xə cʼamma y-asyə-xəma ə-xʸɨr. father-2smPOSS shoe [REL.]3smS-sell.IPFV-COMP 1sS-know.IPFV ʽI know that your father sells shoes.ʼ (941) za ɨxa-ta tɨ-ʃə-n-xəma an-xar-xʷ. DEM water-3smPOSS [REL.]3sfS-want.IPFV-3smO-COMP NEG-know.PFV-1sS ‘I did not know that she wants this water.’

To express ʽwhether (or not)ʼ, the verb in the complement clause is repeated and negated, with the complementizer suffixed to the second, negated verb only (942).

(942) yə-mʷət-ə am-mʷət-ə-xəma e-xʸr-i. REL-die.PFV-3smS [REL.]NEG-die.PFV-3smS-COMP NEG.3S-know.IPS.IPFV-3smO ‘It is not know whether he is dead (or not).’

256 4.7.3.5 Conditional clauses Real conditional clauses are formed with bə- +  (943). The verb of the apodosis can in principle occur in any regular main verb form as for example Imperfective or Definite Future in (944), or the Indefinite Future in (945) .

(943) bə-zənəb-ə ambɨr jɨ-rəma. COND-rain.PFV-3smS cabbage 3smS-grow.IPFV ‘If it rains, the cabbage grows/will grow.’ (944) bə-trəssa-xʷ y-aʒ-e(-te). COND-get.up.PFV-1sS 3smS-see.IPFV-1sO(-DEF.FUT) ʽIf I stand up, he sees me (he will see me).ʼ (945) amir b-aβ-kə tɨ-sd-ɨn-ʃə? A. COND-give.PFV-1sS.2smO 2smS-take.JUS-3smO-INDEF.FUT ʽIf I give you Amir, will/would you take him [with you]?ʼ

Irreal and counterfactual conditional clauses are formed with tə- +  (946)-(947). The verb of the apodosis is a Jussive with the past auxiliary banə ~-ba (↗ 3.18.4). Note that in contrast to the actual Jussive the subject markers correspond to the ones of the Imperfective (i.e. ”archaic jussive”, see table 59).

(946) tə-zənəb-ə ambɨr jɨ-rma-ba. COND-rain.PFV-3smS cabbage 3smS-grow.JUS-AUX.PT ‘If it rained, the cabbage would grow.’~ ‘If it had rained, the cabbage would have grown.’ (947) tə-trəssa-xʷ yɨ-ʒ-e-βa. COND-get.up.PFV-1sS 3smS-see.JUS-1sO-AUX.PT ʽIf I had got up, he would have seen me.ʼ

4.7.3.6 Purposive clauses Purposive clauses (ʽin order toʼ) are formed by -e suffixed to the (relative)  verb (948)-(949).

(948) ə-wərd-e [REL.]1sS-go.down.IPFV-PURP ‘in order for me to go down’ (949) zɨxe tɨ-tmər-e cənə-c-ɨm. here [REL.]3sfS-learn.IPFV-PURP come.PFV-3sfS-M ‘She came here to study.’

After vowels the marker is often realized as -y (950). However, the full form is also possible, either with inserted glide (951) or alternatively forming a hiatus with o and a (952).

257 (950) yɨ-manx-e-y [REL.]3smS-capture.IPFV-1sO-PURP ‘in order for him to capture me’ (951) yɨ-məkʸr-e-ye [REL.]3smS-burn.IPFV-1sO-PURP ‘in order for it to burn me’ (952) yɨ-srəβ-o-e ~yɨ-srəβ-o-ye [REL.]3S-buy.IPFV-pmS-PURP ‘in order for them (m) to buy’

The vowel ə it is deleted when -e is suffixed (953) (cf. example (89)). Deletion is also possible with the vowel a (954), however this output could also be analyzed as the alternative monophthong pronounciation e (~ɛ) of ay (↗ 2.2.1).

(953) yɨ-səmʷə-pʷ-e < *yɨ-səmʷə-pʷə-e [REL.]3S-hear.IPFV.IPS-1sO-MAL.3sm-PURP ‘in order for one to hear with it’ (954) yɨ-gəβ-e < yɨ-gəβa-e or yɨ-gəβa-y [REL.]3smS-enter.IPFV-PURP ‘in order (for him) to enter’

Note that in a sequence of three vowels as it occurs in 3pmS -o + 3smO -i +  -e the medial i has to be pronounced as geminated glide yy (955). If there is only a simple y, this is understood as the epenthetic glide between 3pmS -o +  -e without 3smO -i (956) (cf. example (952) above).

(955) yɨβroyye < /yɨ-βr-o-i-e/ [REL.]3S-say.IPFV-pmS-3smO-PURP ‘in order for them (m) to say him’ (956) yɨβroye < /yɨ-βr-o-e/ [REL.]3S-say.IPFV-pmS-3smO-PURP ‘in order for them (m) to say’

Negated puropsive clauses additionally feature the prefix b- (957)-(958). Thus it is formally identical to a Negated Past Imperfective + -e (↗ 3.15.5) or a temporal verb that is negated + -e, but note that negated temporals are normally formed with t- (↗ 4.7.3.3). Here the gloss  has been chosen.

(957) nɨfas b-e-gəβ-e bərr yɨ-zəgʷe. wind TEMP-[REL.]NEG.3smS-enter.IPFV-PURP door 3smS-close.IPFV.IPS.3smO ʽThe door is shut so that the wind does not come in.ʼ (958) əxɨr b-e-rəma-m-e kʼʷənə-βo-m barley TEMP-[REL.]NEG.3smS-grow.IPFV-ALSO-PURP roast.PFV-3pmS-CV.M zən-əw-i-m. sow.PFV-3pmS-3smO-M ‘In order that the barley does not grow, they roasted and sowed it.’

258 When purposive clauses feature a different subject than the main clause they are usually expressed by the ʽcomplement clauseʼ construction with -xəma (↗ 4.7.3.4). Compare sentence (959) with same subjects to sentence (960) with different subjects. The negated purposive clause in (961) lacks the prefix b- (in contrast to (957)-(958) above), but b-e-wətʼkʼ-ɨwə-xəma is equally possible.

(959) kʼawa afətʼr-e ɨsat məkkʸər-xʷ-ɨm. coffee [REL.][1sS.]boil.IPFV-PURP fire set.fire.PFV-1sS-M ‘I made a fire to boil coffee.’ (960) adot-əna kʼawa t-afətʼɨr-xəma ɨsat mother-1sPOSS coffee [REL.]3sfS-boil.IPFV-COMP/PURP fire məkkʸər-xɨ-la-m. set.fire.PFV-1sS-BEN.3sf-M ‘I made a fire in order for my mother to make coffee.’ (961) yə-fərəz-əta əcʼə e-wətʼkʼ-ɨwə-xəma DAT-horse-3smPOSS wood [REL.]NEG.3smS-fall.IPFV-MAL.3sm-COMP/PURP gafərar akəna-m agʷəd-ə-n-ɨm. part.inside.house ascend.PFV[.3smS]-CV.M tie.PFV-3smS-3smO-M ‘In order that no wood fall on his horse, he brought it to the gafərar and tied it.’

Purposive clauses occur regularly as complements of the verbs ʃə ʽwantʼ (962) and xəna ʽpreventʼ (963).²¹⁸

(962) tɨ-kʼətʼɨn-no-e ʃə-c-ɨm. [REL.]3sfS-kill.IPFV-3pmO-PURP want.PFV-3sfS-M ʽShe wanted to kill them (m)ʼ. (963) yɨ-rəda-na-y mɨr yɨ-xʷəra-n? [REL.]3smS-help.IPFV-3sfO-PURP what 3smS-prevent.IPFV-3smS ‘What does prevent him from helping her?’

Finally, for purposive readings of Jussive in pseudo direct speech plus converb of bar ʽsayʼ as subordinator, see section 3.17.3.

4.7.3.7 Causal clauses Causal clauses (ʽbecauseʼ) are expressed with the ʽrelativeʼ  and suffixed -e () (964), or by a ʽrelative verbʼ form followed by yəxəre, i.e. the ʽrelativeʼ  of xər ʽbecomeʼ plus the same marker -e (965). Note that for  causals, only the latter form is possible (966);  with -e can only be understood as purposive (↗ 4.7.3.6). In contrast,  causals with and without yəxəre seem to be interchangeable.

²¹⁸Further purposive clauses are also used with the Amharic loans cal ʽcanʼ and fəkkʼəd ʽallowʼ

259 (964) tɨrama bay yə-bar-e-ye a-n-nəmʷd-ɨn. yesterday no REL-say.PFV-3smS.1sO-PURP NEG-1sS-love.IPFV-3smO ʽI donʼt like him because he said no to me yesterday.ʼ (965) gaʃʃe-na yə-kʼʸəmmʷə-n yə-xər-e father-1sPOSS REL-fall.ill.PFV[.3smS]-3smO REL-become.PFV-3smS.PURP ʽbecause my father was sickʼ (966) bə-gzɨyə yɨ-məʃ yə-xər-e nɨ-dəppɨn-nə. LOC-time 3smS-become.night.IPFV REL-become.PFV-3smS.PURP 1pS-finish.JUS-1pS ʽLets finish because it will become night soon.ʼ

260 References

Amberber, Mengistu (2000). Valency-changing and valency-encoding devices in Amharic. In: R.M.W. Dixon & Alexandra Y. Aikhenvald (eds.), Changing valency. Cambridge: CUP, 312-332.

Banksira, Degif Petros (1999a). Du rôle des préfixes a-, tə- et at- en chaha. Recherches Linguistiques / Linguistic Research. Publications de l’Institut d’Etudes et de Recherches pour l’Arabisation 4/1-2, 35-79.

Banksira, Degif Petros (1999b). Chaha subject affixes as two independent heads. In: Karlos Arregi, Benjamin Bruening, Cornelia Krause & Vivian Lin (eds.), Papers on morphology and syntax, cycle one. MIT Working Papers in Linguistics 33, 27-45.

Banksira, Degif Petros (2000). Sound mutations: The morphophonology of Chaha. Amsterdam: John Benjamins.

Bertinetto, P. M. (2006). Perfectives, imperfectives, and progressives. In: Keith Brown (ed.), Encyplopedia of Language & Linguistics (Second Edition). Vol. 9. Oxford: Elsevier, 266-270.

Bohnemeyer, Jürgen, Melissa Bowerman & Penelope Brown (2001). Cut and break clips. In: Stephen C. Levinson & N.J. Enfield (eds.), Manual for the field season 2001. Nijmegen: Max Planck Institute for Psycholinguistics, 90-96.

Bowerman, Melissa, Marianne Gullberg, Asifa Majid & Bhuvana Narasimhan (2004). Put project: the cross-linguistic encoding of placement events. In: Asifa Majid (ed.), Field Manual Volume 9. Nijmegen: Max Planck Institute for Psycholinguis- tics, 10-24.

Cohen, David, Marie-Claude Simeone-Senelle & Martine Vanhove (2002). The grammaticalization of ‘say’ and ‘do’: An areal phenomenon in East Africa. In: Tom Güldemann & Manfred von Roncador (eds.), Reported discourse: A meeting ground for different linguistic domains. Amsterdam: Benjamins, 227-251.

Comrie, Bernard (1976). Aspect. Cambridge: CUP.

Chiarini, G. (1887). Note grammaticali e vocaboli della lingua ciahà (guraghè). In: Antonio Checchi, Da Zeila alle frontiere del Caffa. Vol. III. Roma: Ermano Loescher & Co., 469-484.

Crass, Joachim & Ronny Meyer (2007). Deictics, Copula and Focus in the Ethiopian Convergence Area. Köln: Köppe.

261 Crass, Joachim & Ronny Meyer (2008). Ethiopia. In: Bernd Heine & Derek Nurse (eds.), A linguistic geography of Africa. Cambridge: Cambridge University Press, 228-250.

Dixon, R.M.W. & Alexandra Y. Aikhenvald (eds.) (2000). Changing valency. Cambridge: CUP.

Federal Republic of Ethiopia (2008). Summary and Statistical Report of the 2007 Population and Housing Census. Addis Ababa: Federal Republic of Ethiopia - Population Census Commission.

Ford, Carolyn (1991). Notes on the phonology and grammar of Chaha-Gurage. Journal of Afroasiatic Languages 3, 231-296.

Ford, Carolyn (2003). Chaha language. In: Siegbert Uhlig (ed.), Encyclopaedia Aethiopica. Vol. 1 (A-C). Wiesbaden: Harrassowitz, 663-664.

Fraurud, Kari (2001). Possessives with extensive use: A source of definite articles? In: Irène Baron, Michael Herslund & Finn Sørensen (eds.), Dimensions of Possessions. Amster- dam: Benjamins, 243-267.

FRE see Federal Republic of Ethiopia

ገብረኢየሱስ, ኃ/ማርያም [Gabreyesus, Hailemariam] (1960 E.C.) [1967/68]. የጫሙት ሽካ [yəc’amut ʃɨka]. Addis Ababa: Bɨrhanɨnna Səlam Printing Press.

Goldenberg, Gideon (2005). Gurage. In: Siegbert Uhlig (ed.), Encyclopaedia Aethiopica. Vol. 2 (D-Ha). Wies- baden: Harrassowitz, 924-928.

Gósy, Mária (2004). The manifold function of schwa. Grazer Linguistische Studien 62, 15-26.

Haspelmath, Martin (1997). Indefinite Pronouns. Oxford: Oxford University Press.

Hetzron, Robert (1972). Ethiopian Semitic. Studies in classification. Manchester: Manchester University Press.

Hetzron, Robert (1977). The Gunnän-Gurage languages. Napoli: Istituto Orientale di Napoli.

Hetzron, Robert (1996). The two futures in Central and Peripheral Western Gurage. In: Grover Hudson

262 (ed.), Essays on Gurage language and culture. Dedicated to Wolf Leslau on the occasion of his 90th birthday. Wiesbaden: Harrassowitz, 101-109.

Hetzron, Robert (1997). Outer South Ethiopic. In: Robert Hetzron (ed.), The Semitic languages. London: Routledge, 535-549.

Hudson, Grover (1985). The principled grammar of Amharic verb stems. Journal of African Languages and Linguistics 7.1, 39-85.

International Phonetic Association (1999). Handbook of the International Phonetic Association. Cambridge: Cambridge Uni- versity Press.

Kenstowicz, Michael & Degif Petros Banksira (1999). Reduplicative identity in Chaha. Linguistic Inquiry 30, 573-585.

Kulikov, Leonid I. (2001). Causatives. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.), Language Typology and Language Universals. HSK 20:2. Berlin: Walter de Gruyter, 886-898.

Ladefoged, Peter (1996). Elements of acoustic phonetics. 2nd edition. Chicago: The University of Chicago Press.

Leslau, Wolf (1948). Le problème de la gémination du verbe tchaha (gouragué). Word 4, 42-47.

Leslau, Wolf (1950). Ethiopic documents: Gurage. New York: Viking Fund.

Leslau, Wolf (1964). The Jussive in Chaha. Language 40: 53-57.

Leslau, Wolf (1966). Ethiopians Speak, Studies in Cultural Background, II. Chaha. Berkeley & Los An- geles: University of California Press.

Leslau, Wolf (1967). The impersonal in Chaha. In: To honor Roman Jakobson: Essays on the occasion of his seventieth birthday. Vol. 2. The Hague: Mouton, 1150-1162.

Leslau, Wolf (1969). Toward a classification of the Gurage dialects. Journal of Semitic Studies 14, 96- 109.

263 Leslau, Wolf (1979a). Etymological dictionary of Gurage (Ethiopia). Vol. I: Individual dictionaries. Wies- baden: Harrassowitz.

Leslau, Wolf (1979b). Etymological dictionary of Gurage (Ethiopia). Vol. II: English-Gurage index. Wies- baden: Harrassowitz.

Leslau, Wolf (1979c). Etymological dictionary of Gurage (Ethiopia). Vol. III: Etymological section. Wies- baden Harrassowitz.

Leslau, Wolf (1981). Ethiopians speak: Studies in cultural background IV, Muher. Wiesbaden: Franz Steiner.

Leslau, Wolf (1983). Ethiopians speak: Studies in cultural background V, Chaha-Ennemor. Wiesbaden: Franz Steiner.

Leslau, Wolf (1987). Comparative dictionary of Geʿez (Classical Ethiopic). Wiesbaden: Harrassowitz.

Leslau, Wolf (1992). The pseudo-gerundive in Chaha. In: Wolf Leslau, Gurage Studies: Collected Ar- ticles. Wiesbaden: Harrassowitz, 443-458. (Originally published 1969 in Rassegna di studi etiopici 23, 27-42.)

Leslau, Wolf (1995). Reference grammar of Amharic. Wiesbaden: Harrassowitz.

Leslau, Wolf (1996). Čaha and Ennemor: An analysis of two Gurage dialects. In: Grover Hudson (ed.), Essays on Gurage language and culture. Wiesbaden: Harrasowitz, 111-122.

Leslau, Wolf (1997). Chaha (Gurage) phonology. In: Alan S. Kaye (ed.), Phonologies of Asia and Africa. Vol. 1. Winona Lake: Eisenbrauns, 373-398.

Lewis, M. Paul (ed.) (2009). Ethnologue: Languages of the World, Sixteenth edition. Dallas, Tex.: SIL Interna- tional. Online version: http://www.ethnologue.com/.

Lowenstamm, Jean (1996). Five puzzling Chaha verbs: an exercise in practical morphophonemics. In: Grover Hudson (ed.), Essays on Gurage Language and Culture. Wiesbaden: Har- rassowitz, 123-132.

264 Lowenstamm, Jean (2000). The No straddling Effect and its Interpretation: A Fromal Property of Chaha 2nd Feminine Singular Formation. In: Jacqueline Lecarme, Jean Lowenstamm & Ur Shlonsky (eds.), Research in Afroasiatic Grammar. Papers from the third conference on Afroasiatic languages, Sophia Antipolis, France, 1996. Amsterdam: Benjamins, 183-198. Lyons, Christopher (1999). Definiteness. Cambridge: CUP. McCarthy, John (1983). Consonantal morphology in the Chaha verb. Proceedings of WCCFL 2: 176-188. McCarthy, John (1986). Lexical phonology and nonconcatenative morphology in the history of Chaha. Revue québécoise de linguistique 16.1, 209-228. Menuta, Fekede (2002). Morphology of Eža. M.A. thesis, Addis Abeba University. Messele, Abebayehu (2007). An acoustic analysis of a pathological speech: the case of an Amharic speaking person with flaccid dysarthria. M.A. thesis, Addis Ababa University. Meyer, Ronny (2005a). Gumär language. In: Siegbert Uhlig (ed.), Encyclopaedia Aethiopica. Vol. 2 (D- Ha). Wiesbaden: Harrassowitz, 912-914. Meyer, Ronny (2005b). Das Zay. Deskriptive Grammatik einer Ostguragesprache (Äthiosemitisch). Köln: Köppe. Meyer, Ronny (2006). Wolane. Descriptive grammar of an East Gurage language (Ethiosemitic). Köln: Köppe. Meyer, Ronny (2007). Non-verbal predication in East Gurage and Gunnän Gurage languages. In: Joachim Crass & Ronny Meyer (eds.), Deictics, Copula and Focus in the Ethiopian Convergence Area. Köln: Köppe, 177-194. Meyer, Ronny (2009). The quotative verb in Ethiosemitic languages and in Oromo. In: Joachim Crass & Ronny Meyer (eds.), Language Contact and Language Change in Ethiopia. (Topics in African Studies 14). Köln, Köppe, 17-42. Meyer, Ronny (2011). Gurage. In: Stefan Weninger (ed.), The Semitic Languages. An International Handbook. HSK 36. Berlin: De Gruyter Mouton, 1220-1257.

265 Nishi, Makato (2005). Making and unmaking of the nation-state and ethnicity in modern Ethiopia: a study on the history of the Silte people. African Study Monographs. Suppl. 29. Kyoto: The Center for African Area Studies, 157-168.

Oostendorp, Marc van (1999). Schwa in Phonological Theory. Online: www.vanoostendorp.nl/fonologie/ schwaip.htm (accessed April 2009)

Payne, Thomas (1997). Describing morphosyntax: A guide for field linguists. Cambridge: CUP. Petros, Degif (1993). La dérivation verbale en chaha. M.A. thesis, Université du Québec à Montréal. Petros, Degif (1994). On Prefix-Necessitating Stems in Chaha. In: Harold Marcus & Grover Hudson (eds.), New Trends in Ethiopian Studies. Proceedings of the 12th International Con- ference of Ethiopian Studies. Lawrenceville: Red Sea Press, 1220-1235. Petros, Degif (1996a). Sonorant Alternations in Chaha. In: Grover Hudson (ed.), Essays on Gurage Lan- guage and Culture. Wiesbaden: Harrassowitz, 153-173. Petros, Degif (1996b). Sur l’alternance inaccusative vs. inergative en Chaha. Cahiers de linguistique de l’UQAM 1, 71-87. Petros, Degif (1996c). On the absence of AgrS: Evidence from Ethiopian Semitic languages. In: Anna- Maria Di Sciullo (ed.), Configurations. Somerville: Cascadilla Press, 129-160. Polotsky, Hans Jakob. (1951) Notes on Gurage grammar. Oriental Notes and Studies 2. Jerusalem: Israel Ori- ental Society.

Prunet, Jean-François & Degif Petros (1996). L’interaction entre schèmes et racines en chaha. In: Jacqueline Lecarme, Jean Lowenstamm & Ur Shlonsky (eds.), Studies in Afroasiatic grammar. The Hague: Holland Academic Graphics, 302-336.

Rose, Sharon (1994). Palatalization, Underspecification and Plane Conflation in Chaha. In: Eric Dun- can, Donka Farkas & Philip Spaelti (eds.), Proceedings of the 12th West Coast Conference on Formal Linguistics. Stanford (CA): CSLI, 101-116. Rose, Sharon (2000). Epenthesis Positioning and Syllable Contact in Chaha. Phonology 17.3, 397-425.

266 Rose, Sharon (2006). Durational conditions on Endegeň gemination. In: Siegbert Uhlig (ed.), Proceed- ings of the XVth International Conference of Ethiopian Studies. Wiesbaden: Har- rassowitz, 843-850.

Rose, Sharon (2007). Chaha (Gurage) Morphology. In: Alan S. Kaye (ed.), Morphologies of Asia and Africa. Vol. 1. Winona Lake (IN): Eisenbrauns, 403-427.

Shack, William (1966). The Gurage: A people of the Ensete culture. London: Oxford University Press.

Shack, William & Habte-Mariam Marcos (1974). Gods and Heroes: Oral Traditions of the Gurage of Ethiopia. Oxford: Clarendon Press.

Ueno, Mieko (2004). On at-causatives of transitive verbs in Chaha. In: Andrew Simpson (ed.), Pro- ceedings of the Twenty-Seventh Annual Meeting of the Berkeley Linguistics Soci- ety. Special Session on Afroasiatic Languages. Berkeley (CA): BLS, 109-121.

Ullendorff, Edward (1955). The Semitic Languages of Ethiopia. A Comparative Phonology. London: Taylor’s (Foreign) Press.

Völlmin, Sascha (2009). Some dialectal differences between Gumer and Chaha (Gurage). In: Joachim Crass & Ronny Meyer (eds.), Language Contact and Language Change in Ethiopia. Topics in African Studies 14. Köln: Köppe, 83-95.

Völlmin, Sascha (2010a). Benefactives and malefactives in Gumer (Gurage). In: Fernando Zúñiga & Seppo Kittilä (eds.), Benefactives and malefactives: Typological perspectives and case studies. Typological Studies in Language 92. Amsterdam: Benjamins, 317-330.

Völlmin, Sascha (2010b). The so-called converbs in Gumer (Gurage). In: Sascha Völlmin, Azeb Amha, Christian Rapold & Silvia Zaugg-Coretti (eds.), Converbs, medial verbs, clause chaining and related issues. Frankfurter Afrikanistische Blätter 19 [2007]. Köln: Köppe, 81-97.

Wetter, Andreas (2010). Das Argobba. Eine deskriptive Grammatik der Varietät von Shonke und T’ollaha. Köln: Köppe.

Zaugg-Coretti, Silvia (2009). The morpheme -tu as a focus marker in Yemsa (Omotic) and Oromo (Cushitic).

267 In: Joachim Crass & Ronny Meyer (eds.), Language Contact and Language Change in Ethiopia. Topics in African Studies 14. Köln: Köppe, 97-120.

268